本公开涉及计算机技术领域，尤其涉及一种任务处理方法及装置、电子设备和存储介质。The present disclosure relates to the field of computer technology, and in particular to a task processing method and device, an electronic device, and a storage medium.
背景技术Background Art
连续学习指人工智能模型或系统(例如，神经网络)顺序的学习一系列任务，并且在学习新任务时不能使用已学任务的数据。这种学习范式适合数据连续不断输入的实际应用场景，是通用人工智能的基本要求。实现连续学习的关键是学得的新知识不能覆盖已学的旧知识，从而避免发生灾难性遗忘(catastrophic forgetting)。Continuous learning refers to an AI model or system (e.g., a neural network) sequentially learning a series of tasks, and cannot use data from previously learned tasks when learning new tasks. This learning paradigm is suitable for practical application scenarios where data is continuously input, and is a basic requirement for general artificial intelligence. The key to achieving continuous learning is that new knowledge cannot overwrite old knowledge, thereby avoiding catastrophic forgetting.
在相关技术中，就神经网络的连续学习算法而言，一种方法是通过一定的正则化项来约束网络的参数，使神经网络在学习任务的同时，也能保持在已学任务上的良好性能。弹性权重固化(elastic weight consolidation)算法是这种方法的代表，它通过选择性的降低对已学任务重要的权重的学习率来克服灾难性遗忘。另外一种是通过一定机制为不同的任务动态分配不同的参数，使不同任务的参数减少耦合，从而避免重要参数在学习中被覆盖。场景依赖门控(context-dependent gating)算法是这种方法的代表，它通过在处理不同任务时随机掩蔽部分神经元来缓解学习新任务时对网络参数的干扰，减缓遗忘。场景依赖门控算法为不同任务分配一个随机的二值掩蔽向量，来控制隐层神经元的激活状态。在学习一个特定任务时，将根据该任务对应的掩蔽向量来关闭网络的部分神经元，从而避免与之相关的参数在学习这个任务时被修改。这相当于用不相交参数集合来处理不同的任务，减少参数间的相互干扰，缓解灾难性遗忘。然而，该方法只缓解了不相关任务间的参数干扰，没有利用任务相关性来复用网络参数，从而提高网络参数的利用效率。In the related art, as far as the continuous learning algorithm of the neural network is concerned, one method is to constrain the parameters of the network through a certain regularization term so that the neural network can maintain good performance on the learned task while learning the task. The elastic weight consolidation algorithm is a representative of this method, which overcomes catastrophic forgetting by selectively reducing the learning rate of the weights that are important to the learned task. Another method is to dynamically assign different parameters to different tasks through a certain mechanism to reduce the coupling of the parameters of different tasks, thereby avoiding important parameters from being covered during learning. The context-dependent gating algorithm is a representative of this method, which alleviates the interference of network parameters when learning new tasks and slows down forgetting by randomly masking some neurons when processing different tasks. The context-dependent gating algorithm assigns a random binary masking vector to different tasks to control the activation state of the hidden layer neurons. When learning a specific task, some neurons of the network will be turned off according to the masking vector corresponding to the task, thereby avoiding the modification of the parameters related to it when learning the task. This is equivalent to using disjoint parameter sets to process different tasks, reducing mutual interference between parameters, and alleviating catastrophic forgetting. However, this method only alleviates the parameter interference between unrelated tasks, and does not utilize task correlation to reuse network parameters to improve the utilization efficiency of network parameters.
发明内容Summary of the invention
本公开提出了一种任务处理方法及装置、电子设备和存储介质。The present disclosure proposes a task processing method and device, an electronic device and a storage medium.
根据本公开的一方面，提供了一种任务处理方法，包括：将目标任务的待处理信息输入调制网络，获得与所述目标任务对应的调制信息，所述目标任务是多个预设任务中的任意一个；根据所述调制信息，从任务网络的多个网络节点中，确定出用于处理所述待处理信息的目标网络节点；通过所述目标网络节点对所述待处理信息进行处理，获得所述目标任务的处理结果。According to one aspect of the present disclosure, a task processing method is provided, including: inputting information to be processed of a target task into a modulation network to obtain modulation information corresponding to the target task, wherein the target task is any one of a plurality of preset tasks; according to the modulation information, determining a target network node for processing the information to be processed from a plurality of network nodes of a task network; processing the information to be processed by the target network node to obtain a processing result of the target task.
在一种可能的实现方式中，所述预设任务包括第一任务和第二任务，所述第一任务对应的调制信息与所述第二任务对应的调制信息的信息相似度，与所述第一任务和所述第二任务的任务相似度正相关。In a possible implementation, the preset task includes a first task and a second task, and information similarity between modulation information corresponding to the first task and modulation information corresponding to the second task is positively correlated with task similarity between the first task and the second task.
在一种可能的实现方式中，所述预设任务包括第一任务和第二任务，所述第一任务对应的调制信息与所述第二任务对应的调制信息的信息相似度，与所述任务网络中的第一网络节点和第二网络节点中网络节点的重复率正相关，其中，所述第一网络节点为所述任务网络中用于处理第一任务的待处理信息的目标网络节点，所述第二网络节点为所述任务网络中用于处理第二任务的待处理信息的目标网络节点。In a possible implementation, the preset task includes a first task and a second task, and the information similarity between the modulation information corresponding to the first task and the modulation information corresponding to the second task is positively correlated with the repetition rate of the network nodes in the first network node and the second network node in the task network, wherein the first network node is the target network node in the task network for processing the information to be processed of the first task, and the second network node is the target network node in the task network for processing the information to be processed of the second task.
在一种可能的实现方式中，所述调制网络包括人工神经网络和脉冲神经网络中的任意一种，所述任务网络包括脉冲神经网络和人工神经网络中的任意一种。In a possible implementation, the modulation network includes any one of an artificial neural network and a spiking neural network, and the task network includes any one of a spiking neural network and an artificial neural network.
在一种可能的实现方式中，所述方法还包括：将训练任务的训练样本输入所述调制网络，获得所述训练样本的第一训练调制信息，所述训练任务是多个任务中的任意一个；根据所述训练任务的多个训练样本的第一训练调制信息，获得所述训练任务的第二训练调制信息；根据所述第一训练调制信息和所述第二训练调制信息，训练所述调制网络。In a possible implementation, the method further includes: inputting a training sample of a training task into the modulation network to obtain first training modulation information of the training sample, wherein the training task is any one of a plurality of tasks; obtaining second training modulation information of the training task based on the first training modulation information of the plurality of training samples of the training task; and training the modulation network based on the first training modulation information and the second training modulation information.
在一种可能的实现方式中，根据所述训练任务的多个训练样本的第一训练调制信息，获得所述训练任务的第二训练调制信息，包括：获取所述多个训练样本的第一训练调制信息的平均训练调制信息；对所述平均训练调制信息进行阈值化处理，获得所述第二训练调制信息。In a possible implementation, obtaining second training modulation information of the training task based on first training modulation information of multiple training samples of the training task includes: obtaining average training modulation information of the first training modulation information of the multiple training samples; and performing thresholding processing on the average training modulation information to obtain the second training modulation information.
在一种可能的实现方式中，根据所述第一训练调制信息和所述第二训练调制信息，训练所述调制网络，包括：根据所述第一训练调制信息和所述第二训练调制信息，获得相似度约束信息；根据所述第二训练调制信息，获得正则化信息；根据所述相似度约束信息和所述正则化信息，确定所述调制网络的第一网络损失；根据所述第一网络损失，训练所述调制网络。In a possible implementation, the modulation network is trained according to the first training modulation information and the second training modulation information, including: obtaining similarity constraint information according to the first training modulation information and the second training modulation information; obtaining regularization information according to the second training modulation information; determining a first network loss of the modulation network according to the similarity constraint information and the regularization information; and training the modulation network according to the first network loss.
在一种可能的实现方式中，所述方法还包括：根据训练任务的训练样本以及已训练的调制网络，获得所述训练任务的第三训练调制信息，所述训练任务是多个任务中的任意一个；根据所述第三训练调制信息确定所述任务网络中用于处理所述训练任务的训练样本的目标网络节点；通过所述目标网络节点处理所述训练任务的训练样本，获得训练结果；根据所述训练结果与所述训练样本的标注信息确定所述任务网络的第二网络损失；根据所述第二网络损失，训练所述任务网络。In a possible implementation, the method further includes: obtaining third training modulation information of the training task based on the training samples of the training task and a trained modulation network, wherein the training task is any one of multiple tasks; determining a target network node in the task network for processing the training samples of the training task based on the third training modulation information; processing the training samples of the training task through the target network node to obtain a training result; determining a second network loss of the task network based on the training result and the labeling information of the training samples; and training the task network based on the second network loss.
在一种可能的实现方式中，根据训练任务的训练样本以及已训练的调制网络，获得所述训练任务的第三训练调制信息，包括：在所述任务网络为单隐层神经网络的情况下，根据所述调制网络获得的当前训练任务的第四训练调制信息，以及已训练的历史训练任务的第五训练调制信息，确定当前训练任务的第三训练调制信息。In one possible implementation, third training modulation information of the training task is obtained based on the training samples of the training task and the trained modulation network, including: when the task network is a single hidden layer neural network, the third training modulation information of the current training task is determined based on the fourth training modulation information of the current training task obtained from the modulation network and the fifth training modulation information of the trained historical training task.
在一种可能的实现方式中，所述多个预设任务包括图像处理任务、语音处理任务、文字处理任务、向量处理任务中的至少一种，所述待处理信息包括图像信息、语音信息、文字信息、向量信息中的至少一种。In a possible implementation, the multiple preset tasks include at least one of an image processing task, a voice processing task, a text processing task, and a vector processing task, and the information to be processed includes at least one of image information, voice information, text information, and vector information.
根据本公开的一方面，提供了一种任务处理装置，包括：调制信息模块，用于将目标任务的待处理信息输入调制网络，获得与所述目标任务对应的调制信息，所述目标任务是多个预设任务中的任意一个；目标节点模块，用于根据所述调制信息，从任务网络的多个网络节点中，确定出用于处理所述待处理信息的目标网络节点；处理模块，用于通过所述目标网络节点对所述待处理信息进行处理，获得所述目标任务的处理结果。According to one aspect of the present disclosure, a task processing device is provided, including: a modulation information module, used to input the information to be processed of the target task into a modulation network to obtain modulation information corresponding to the target task, wherein the target task is any one of a plurality of preset tasks; a target node module, used to determine, according to the modulation information, a target network node for processing the information to be processed from a plurality of network nodes of the task network; and a processing module, used to process the information to be processed through the target network node to obtain a processing result of the target task.
在一种可能的实现方式中，所述预设任务包括第一任务和第二任务，所述第一任务对应的调制信息与所述第二任务对应的调制信息的信息相似度，与所述第一任务和所述第二任务的任务相似度正相关。In a possible implementation, the preset task includes a first task and a second task, and information similarity between modulation information corresponding to the first task and modulation information corresponding to the second task is positively correlated with task similarity between the first task and the second task.
在一种可能的实现方式中，所述预设任务包括第一任务和第二任务，所述第一任务对应的调制信息与所述第二任务对应的调制信息的信息相似度，与所述任务网络中的第一网络节点和第二网络节点中网络节点的重复率正相关，其中，所述第一网络节点为所述任务网络中用于处理第一任务的待处理信息的目标网络节点，所述第二网络节点为所述任务网络中用于处理第二任务的待处理信息的目标网络节点。In a possible implementation, the preset task includes a first task and a second task, and the information similarity between the modulation information corresponding to the first task and the modulation information corresponding to the second task is positively correlated with the repetition rate of the network nodes in the first network node and the second network node in the task network, wherein the first network node is the target network node in the task network for processing the information to be processed of the first task, and the second network node is the target network node in the task network for processing the information to be processed of the second task.
在一种可能的实现方式中，所述调制网络包括人工神经网络和脉冲神经网络中的任意一种，所述任务网络包括脉冲神经网络和人工神经网络中的任意一种。In a possible implementation, the modulation network includes any one of an artificial neural network and a spiking neural network, and the task network includes any one of a spiking neural network and an artificial neural network.
在一种可能的实现方式中，所述装置还包括调制网络训练模块，所述调制网络训练模块用于：将训练任务的训练样本输入所述调制网络，获得所述训练样本的第一训练调制信息，所述训练任务是多个任务中的任意一个；根据所述训练任务的多个训练样本的第一训练调制信息，获得所述训练任务的第二训练调制信息；根据所述第一训练调制信息和所述第二训练调制信息，训练所述调制网络。In a possible implementation, the device also includes a modulation network training module, which is used to: input a training sample of a training task into the modulation network to obtain first training modulation information of the training sample, where the training task is any one of multiple tasks; obtain second training modulation information of the training task based on the first training modulation information of multiple training samples of the training task; and train the modulation network based on the first training modulation information and the second training modulation information.
在一种可能的实现方式中，所述调制网络训练模块进一步用于：获取所述多个训练样本的第一训练调制信息的平均训练调制信息；对所述平均训练调制信息进行阈值化处理，获得所述第二训练调制信息。In a possible implementation, the modulation network training module is further used to: obtain average training modulation information of first training modulation information of the multiple training samples; and perform thresholding processing on the average training modulation information to obtain the second training modulation information.
在一种可能的实现方式中，所述调制网络训练模块进一步用于：根据所述第一训练调制信息和所述第二训练调制信息，获得相似度约束信息；根据所述第二训练调制信息，获得正则化信息；根据所述相似度约束信息和所述正则化信息，确定所述调制网络的第一网络损失；根据所述第一网络损失，训练所述调制网络。In a possible implementation, the modulation network training module is further used to: obtain similarity constraint information based on the first training modulation information and the second training modulation information; obtain regularization information based on the second training modulation information; determine the first network loss of the modulation network based on the similarity constraint information and the regularization information; and train the modulation network based on the first network loss.
在一种可能的实现方式中，所述装置还包括任务网络训练模块，所述任务网络训练模块用于：根据训练任务的训练样本以及已训练的调制网络，获得所述训练任务的第三训练调制信息，所述训练任务是多个任务中的任意一个；根据所述第三训练调制信息确定所述任务网络中用于处理所述训练任务的训练样本的目标网络节点；通过所述目标网络节点处理所述训练任务的训练样本，获得训练结果；根据所述训练结果与所述训练样本的标注信息确定所述任务网络的第二网络损失；根据所述第二网络损失，训练所述任务网络。In a possible implementation, the device also includes a task network training module, which is used to: obtain third training modulation information of the training task based on the training samples of the training task and a trained modulation network, where the training task is any one of multiple tasks; determine a target network node in the task network for processing the training samples of the training task based on the third training modulation information; process the training samples of the training task through the target network node to obtain a training result; determine a second network loss of the task network based on the training result and the labeling information of the training samples; and train the task network based on the second network loss.
在一种可能的实现方式中，所述任务网络训练模块进一步用于：在所述任务网络为单隐层神经网络的情况下，根据所述调制网络获得的当前训练任务的第四训练调制信息，以及已训练的历史训练任务的第五训练调制信息，确定当前训练任务的第三训练调制信息。In one possible implementation, the task network training module is further used to: when the task network is a single hidden layer neural network, determine the third training modulation information of the current training task based on the fourth training modulation information of the current training task obtained by the modulation network and the fifth training modulation information of the trained historical training tasks.
在一种可能的实现方式中，所述多个预设任务包括图像处理任务、语音处理任务、文字处理任务、向量处理任务中的至少一种，所述待处理信息包括图像信息、语音信息、文字信息、向量信息中的至少一种。In a possible implementation, the multiple preset tasks include at least one of an image processing task, a voice processing task, a text processing task, and a vector processing task, and the information to be processed includes at least one of image information, voice information, text information, and vector information.
根据本公开的一方面，提供了一种电子设备，包括：处理器；用于存储处理器可执行指令的存储器；其中，所述处理器被配置为：执行上述任务处理方法。According to one aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to: execute the above-mentioned task processing method.
根据本公开的一方面，提供了一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述任务处理方法。According to one aspect of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the above-mentioned task processing method is implemented.
根据本公开的实施例的任务处理方法，可通过调制网络生成的调制信息对任务网络中用于处理任务的网络节点进行选择，可减少参数间的相互干扰，缓解灾难性遗忘，并通过调制信息可提高相似任务中网络节点的复用率，提高网络参数的利用效率。并且，可通过脉冲神经网络处理作为任务网络，处理复杂的时序任务，可使脉冲神经网络在调制信号的作用下进行连续多个任务的训练或执行多个任务，提高了神经网络的训练和执行效率。进一步地，在训练调制网络的过程中，可通过相似度约束信息，使调制网络输出相同任务的调制信息相似度较高，在特征空间中更加集中，并使调制网络输出不同任务的调制信息相似度不高，在特征空间中相互远离。使得任务网络通过调制网络输出的调制信息选择目标网络节点时，针对相似的任务选择的目标网络节点重复率较高，针对不相似的任务选择的目标网络节点的重复率较低，可提升网络参数的利用效率，提升任务网络的训练效率，并减少不同任务之间的参数干扰。According to the task processing method of the embodiment of the present disclosure, the modulation information generated by the modulation network can be used to select the network nodes used to process the task in the task network, which can reduce the mutual interference between parameters, alleviate catastrophic forgetting, and improve the reuse rate of network nodes in similar tasks through the modulation information, thereby improving the utilization efficiency of network parameters. In addition, the pulse neural network can be used as a task network to process complex sequential tasks, so that the pulse neural network can be trained or execute multiple tasks continuously under the action of the modulation signal, thereby improving the training and execution efficiency of the neural network. Further, in the process of training the modulation network, the modulation information output by the modulation network for the same task can be made to have a higher similarity and be more concentrated in the feature space through the similarity constraint information, and the modulation information output by the modulation network for different tasks can be made to have a lower similarity and be far away from each other in the feature space. When the task network selects the target network node through the modulation information output by the modulation network, the repetition rate of the target network node selected for similar tasks is higher, and the repetition rate of the target network node selected for dissimilar tasks is lower, thereby improving the utilization efficiency of network parameters, improving the training efficiency of the task network, and reducing the parameter interference between different tasks.
应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，而非限制本公开。It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
根据下面参考附图对示例性实施例的详细说明，本公开的其它特征及方面将变得清楚。Further features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings.
附图说明BRIEF DESCRIPTION OF THE DRAWINGS
此处的附图被并入说明书中并构成本说明书的一部分，这些附图示出了符合本公开的实施例，并与说明书一起用于说明本公开的技术方案。The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used to illustrate the technical solutions of the present disclosure together with the specification.
图1示出根据本公开实施例的任务处理方法的流程图；FIG1 is a flowchart showing a task processing method according to an embodiment of the present disclosure;
图2示出根据本公开实施例的任务处理方法的应用示意图；FIG2 is a schematic diagram showing an application of a task processing method according to an embodiment of the present disclosure;
图3示出根据本公开实施例的任务处理装置的框图；FIG3 shows a block diagram of a task processing device according to an embodiment of the present disclosure;
图4示出根据本公开实施例的任务处理装置的框图；FIG4 shows a block diagram of a task processing device according to an embodiment of the present disclosure;
图5示出根据本公开实施例的任务处理装置的框图。FIG. 5 shows a block diagram of a task processing device according to an embodiment of the present disclosure.
具体实施方式DETAILED DESCRIPTION
以下将参考附图详细说明本公开的各种示例性实施例、特征和方面。附图中相同的附图标记表示功能相同或相似的元件。尽管在附图中示出了实施例的各种方面，但是除非特别指出，不必按比例绘制附图。Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.
在这里专用的词“示例性”意为“用作例子、实施例或说明性”。这里作为“示例性”所说明的任何实施例不必解释为优于或好于其它实施例。The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.
本文中术语“和/或”，仅仅是一种描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本文中术语“至少一种”表示多种中的任意一种或多种中的至少两种的任意组合，例如，包括A、B、C中的至少一种，可以表示包括从A、B和C构成的集合中选择的任意一个或多个元素。The term "and/or" herein is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, A and/or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C can represent including any one or more elements selected from the set consisting of A, B, and C.
另外，为了更好的说明本公开，在下文的具体实施方式中给出了众多的具体细节。本领域技术人员应当理解，没有某些具体细节，本公开同样可以实施。在一些实例中，对于本领域技术人员熟知的方法、手段、元件和电路未作详细描述，以便于凸显本公开的主旨。In addition, in order to better illustrate the present disclosure, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present disclosure can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the main purpose of the present disclosure.
图1示出根据本公开实施例的任务处理方法的流程图，如图1所示，所述方法包括：FIG1 is a flowchart of a task processing method according to an embodiment of the present disclosure. As shown in FIG1 , the method includes:
在步骤S11中，将目标任务的待处理信息输入调制网络，获得与所述目标任务对应的调制信息，所述目标任务是多个预设任务中的任意一个；In step S11, the information to be processed of the target task is input into the modulation network to obtain modulation information corresponding to the target task, where the target task is any one of a plurality of preset tasks;
在步骤S12中，根据所述调制信息，从任务网络的多个网络节点中，确定出用于处理所述待处理信息的目标网络节点；In step S12, according to the modulation information, a target network node for processing the information to be processed is determined from a plurality of network nodes of the task network;
在步骤S13中，通过所述目标网络节点对所述待处理信息进行处理，获得所述目标任务的处理结果。In step S13, the information to be processed is processed by the target network node to obtain a processing result of the target task.
根据本公开的实施例的任务处理方法，可通过调制网络生成的调制信息对任务网络中用于处理任务的网络节点进行选择，可减少参数间的相互干扰，缓解灾难性遗忘，并且通过调制信息可提高相似任务中网络节点的复用率，提高网络参数的利用效率，并提高神经网络的训练效率。According to the task processing method of the embodiment of the present disclosure, the network nodes used to process tasks in the task network can be selected through the modulation information generated by the modulation network, which can reduce mutual interference between parameters and alleviate catastrophic forgetting. In addition, the modulation information can be used to increase the reuse rate of network nodes in similar tasks, improve the utilization efficiency of network parameters, and improve the training efficiency of neural networks.
在一种可能的实现方式中，所述任务处理方法可以由终端设备或其它处理设备执行，其中，终端设备可以为用户设备(User Equipment，UE)、移动设备、用户终端、终端、蜂窝电话、无绳电话、个人数字处理(Personal Digital Assistant，PDA)、手持设备、计算设备、车载设备、可穿戴设备等。其它处理设备可为服务器或云端服务器等。在一些可能的实现方式中，该任务处理方法可以通过处理器调用存储器中存储的计算机可读指令的方式来实现。In a possible implementation, the task processing method may be executed by a terminal device or other processing device, wherein the terminal device may be a user equipment (UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. Other processing devices may be a server or a cloud server, etc. In some possible implementations, the task processing method may be implemented by a processor calling a computer-readable instruction stored in a memory.
在一种可能的实现方式中，通常情况下，神经网络可通过特定的训练，实现特定的功能，例如，可通过带有位置标注的样本图像训练神经网络，使神经网络获得确定图像中目标对象位置的能力，或者，可通过带有轮廓标注的样本图像训练神经网络，使神经网络获得识别图像中目标对象的轮廓的能力。而在更复杂的场景下，需要神经网络同时具有多种能力，并执行多种任务，例如，在监控场景下，神经网络不仅需要检测监控视频中目标对象的位置，还需要识别目标对象的身份，而处理器中存储两个或多个神经网络的架构及参数可能造成存储资源浪费以及处理效率低下等问题，因此，可能需要神经网络能够执行多个任务。In a possible implementation, usually, a neural network can achieve specific functions through specific training. For example, a neural network can be trained through sample images with position annotations to enable the neural network to acquire the ability to determine the position of a target object in an image, or a neural network can be trained through sample images with contour annotations to enable the neural network to acquire the ability to recognize the contour of a target object in an image. In more complex scenarios, neural networks are required to have multiple capabilities and perform multiple tasks at the same time. For example, in a monitoring scenario, a neural network not only needs to detect the position of a target object in a monitoring video, but also needs to identify the identity of the target object. Storing the architecture and parameters of two or more neural networks in a processor may cause problems such as waste of storage resources and low processing efficiency. Therefore, a neural network may be required to be able to perform multiple tasks.
在一种可能的实现方式中，在相关技术中，可在神经网络执行不同的任务，或者针对不同的任务进行训练时，随机屏蔽一部分网络节点，即，训练时不会更新被屏蔽的网络节点的参数，执行任务时不使用被屏蔽的网络节点。可通过这样的方式使神经网络通过不同的节点来执行不同的任务，或者针对不同的任务进行训练。例如，针对任务A，可随机屏蔽一部分节点，并使用未被屏蔽的一组网络节点(例如节点组A)来进行训练或者执行任务A，针对任务B，可随机屏蔽一部分网络节点，并使用未被屏蔽的一组网络节点(例如节点组B)来进行训练或者执行任务B。但由于被屏蔽网络节点是随机选择的，在神经网络训练的过程中，可能针对相似的任务选择不同的网络节点来训练和执行，导致训练效率较低，且网络节点的参数利用效率较低。In a possible implementation, in the related art, a part of the network nodes may be randomly shielded when the neural network performs different tasks or performs training for different tasks, that is, the parameters of the shielded network nodes will not be updated during training, and the shielded network nodes will not be used when performing tasks. In this way, the neural network can perform different tasks through different nodes, or be trained for different tasks. For example, for task A, a part of the nodes may be randomly shielded, and a group of unshielded network nodes (such as node group A) may be used to train or perform task A. For task B, a part of the network nodes may be randomly shielded, and a group of unshielded network nodes (such as node group B) may be used to train or perform task B. However, since the shielded network nodes are randomly selected, during the training of the neural network, different network nodes may be selected for training and execution for similar tasks, resulting in low training efficiency and low parameter utilization efficiency of the network nodes.
在一种可能的实现方式中，针对上述问题，可通过调制网络针对任务的特性来生成调制信号，并通过调制信号来针对性地屏蔽任务网络中的部分网络节点，并使用未被屏蔽的网络节点来进行训练和执行。由于调制信号是根据任务的特性生成的，因此，对于相似的任务，调制信号的相似度较高，基于调制信号选择的用于训练或执行任务的网络节点的重复率也较高，因此，对于相似的任务，可提升神经网络的训练效率以及网络节点的参数利用率。In a possible implementation, in order to solve the above problem, a modulation signal can be generated according to the characteristics of the task through a modulation network, and some network nodes in the task network can be shielded through the modulation signal, and the unshielded network nodes can be used for training and execution. Since the modulation signal is generated according to the characteristics of the task, for similar tasks, the similarity of the modulation signal is high, and the repetition rate of the network nodes selected for training or executing tasks based on the modulation signal is also high. Therefore, for similar tasks, the training efficiency of the neural network and the parameter utilization rate of the network nodes can be improved.
在一种可能的实现方式中，所述多个预设任务包括图像处理任务、语音处理任务、文字处理任务、向量处理任务中的至少一种，所述待处理信息包括图像信息、语音信息、文字信息、向量信息中的至少一种。在示例中，所述任务可以是图像处理任务，例如，对图像中的目标的位置进行识别的任务，则待处理信息可以是图像信息。又例如，所述任务可以是文字处理任务，例如，识别文字的语义，则待处理信息可以是文字信息(例如，词语、短语、句子、文字段落等)。本公开对预设任务和待处理信息的类别不做限制。In a possible implementation, the multiple preset tasks include at least one of an image processing task, a speech processing task, a text processing task, and a vector processing task, and the information to be processed includes at least one of image information, speech information, text information, and vector information. In the example, the task may be an image processing task, such as a task for identifying the position of an object in an image, and the information to be processed may be image information. For another example, the task may be a text processing task, such as recognizing the semantics of text, and the information to be processed may be text information (e.g., words, phrases, sentences, text paragraphs, etc.). The present disclosure does not limit the categories of preset tasks and information to be processed.
在一种可能的实现方式中，在步骤S11中，目标任务可以是多个预设任务中的任意一个，待处理信息可以是与目标任务对应的信息，例如，目标任务是图像处理任务，则待处理信息可以是图像信息。可将目标任务的待处理信息输入调制网络，调制网络可输出与目标任务对应的调制信息。该调制信息用于屏蔽任务网络中的部分网络节点。In a possible implementation, in step S11, the target task may be any one of a plurality of preset tasks, and the information to be processed may be information corresponding to the target task. For example, if the target task is an image processing task, the information to be processed may be image information. The information to be processed of the target task may be input into the modulation network, and the modulation network may output modulation information corresponding to the target task. The modulation information is used to shield some network nodes in the task network.
在示例中，目标任务可以是对图像中的目标对象的位置进行检测的任务，待处理信息可以是图像信息，调制网络可基于输入的图像信息生成该任务的调制信息。又例如，目标任务可以是识别文字的语义的任务，待处理信息可以是文字信息，调制网络可基于输入的文字信息生成该任务的调制信息。目标任务不同，则调制网络生成的调制信息不同。In the example, the target task may be a task of detecting the position of a target object in an image, the information to be processed may be image information, and the modulation network may generate modulation information for the task based on the input image information. For another example, the target task may be a task of recognizing the semantics of text, the information to be processed may be text information, and the modulation network may generate modulation information for the task based on the input text information. Different target tasks generate different modulation information.
在一种可能的实现方式中，所述预设任务包括第一任务和第二任务，所述第一任务对应的调制信息与所述第二任务对应的调制信息的信息相似度，与所述第一任务和所述第二任务的任务相似度正相关。In a possible implementation, the preset task includes a first task and a second task, and information similarity between modulation information corresponding to the first task and modulation information corresponding to the second task is positively correlated with task similarity between the first task and the second task.
在示例中，由于调制信息是基于目标任务生成的，因此，如果两个目标任务相似度较高，则生成的调制信息相似度较高，反之，如果两个目标任务相似度较低，则生成的调制信息相似度较低。例如，第一任务和第二任务均为图像处理任务，第一任务为对图像中目标对象的位置进行识别的任务，第二任务为对图像中目标对象的轮廓进行识别的任务，第一任务和第二任务的相似度较高，则与第一任务对应的调制信息以及与第二任务对应的调制信息的相似度较高。又例如，第一任务为对图像中目标对象的位置进行识别的任务，第二任务为对文字的语义进行识别的任务，第一任务和第二任务的相似度较低，则与第一任务对应的调制信息以及与第二任务对应的调制信息的相似度较低。In the example, since the modulation information is generated based on the target task, if the two target tasks have a high similarity, the generated modulation information has a high similarity. Conversely, if the two target tasks have a low similarity, the generated modulation information has a low similarity. For example, the first task and the second task are both image processing tasks, the first task is a task for identifying the position of a target object in an image, and the second task is a task for identifying the contour of a target object in an image. The similarity between the first task and the second task is high, and the modulation information corresponding to the first task and the modulation information corresponding to the second task have a high similarity. For another example, the first task is a task for identifying the position of a target object in an image, and the second task is a task for identifying the semantics of text. The similarity between the first task and the second task is low, and the modulation information corresponding to the first task and the modulation information corresponding to the second task have a low similarity.
在示例中，调制信息可以是向量形式的信息(例如，调制向量)，在第一任务和第二任务的任务相似度较高时，第一任务对应的调制向量以及第二任务对应的调制向量之间的相似度较高，例如，余弦相似度较高。反之，在第一任务和第二任务的任务相似度较低时，第一任务对应的调制向量以及第二任务对应的调制向量之间的相似度较低，例如，余弦相似度较低。本公开对调制信息的数据类型不做限制。In the example, the modulation information may be information in the form of a vector (e.g., a modulation vector). When the task similarity between the first task and the second task is high, the similarity between the modulation vector corresponding to the first task and the modulation vector corresponding to the second task is high, for example, the cosine similarity is high. Conversely, when the task similarity between the first task and the second task is low, the similarity between the modulation vector corresponding to the first task and the modulation vector corresponding to the second task is low, for example, the cosine similarity is low. The present disclosure does not limit the data type of the modulation information.
在一种可能的实现方式中，所述调制网络包括人工神经网络和脉冲神经网络中的任意一种，所述任务网络包括脉冲神经网络和人工神经网络中的任意一种。在示例中，人工神经网络以时间离散、输出连续为主要特征，适合高精度的数值逼近任务。脉冲神经网络以时间连续、输出离散为主要特征，适合复杂的时序信息处理任务。可将人工神经网络和脉冲神经网络组合成为异构神经网络，并协同完成连续学习以及执行任务。In a possible implementation, the modulation network includes any one of an artificial neural network and a pulse neural network, and the task network includes any one of a pulse neural network and an artificial neural network. In the example, the artificial neural network is characterized by time discreteness and continuous output, and is suitable for high-precision numerical approximation tasks. The pulse neural network is characterized by time continuity and discrete output, and is suitable for complex time series information processing tasks. The artificial neural network and the pulse neural network can be combined into a heterogeneous neural network, and can collaboratively complete continuous learning and execution tasks.
在示例中，可将具有较强拟合能力的人工神经网络作为调制网络，以基于任务的相似度生成调制信息，并将适合复杂的时序信息处理任务的脉冲神经网络作为任务网络，以在调制网络的调制信息的作用下，进行连续多个任务的训练或执行多个任务。本公开对调制网络和任务网络的类型不做限制。例如，调制网络可以是脉冲神经网络，任务网络可以是人工神经网络。或者，调制网络和任务网络均可以是脉冲神经网络，又或者，调制网络和任务网络均可以是人工神经网络。In the example, an artificial neural network with strong fitting ability can be used as a modulation network to generate modulation information based on the similarity of the task, and a pulse neural network suitable for complex time series information processing tasks can be used as a task network to train multiple tasks continuously or execute multiple tasks under the modulation information of the modulation network. The present disclosure does not limit the types of modulation networks and task networks. For example, the modulation network can be a pulse neural network, and the task network can be an artificial neural network. Alternatively, both the modulation network and the task network can be pulse neural networks, or, both the modulation network and the task network can be artificial neural networks.
通过这种方式，可通过脉冲神经网络处理复杂的时序任务，可使脉冲神经网络在调制信号的作用下进行连续多个任务的训练或执行多个任务，提高了神经网络的训练和执行效率。In this way, complex timing tasks can be processed by the pulse neural network, and the pulse neural network can be trained or execute multiple tasks continuously under the action of the modulated signal, thereby improving the training and execution efficiency of the neural network.
在一种可能的实现方式中，在步骤S12中，调制信息可作用于任务网络，控制任务网络屏蔽特定的网络节点，并通过未被屏蔽的网络节点来对待处理信息进行处理，即，将未被屏蔽的网络节点作为用于处理待处理信息的目标网络节点。In one possible implementation, in step S12, the modulation information may act on the task network, control the task network to shield specific network nodes, and process the information to be processed through unshielded network nodes, that is, use the unshielded network nodes as target network nodes for processing the information to be processed.
在一种可能的实现方式中，所述预设任务包括第一任务和第二任务，所述第一任务对应的调制信息与所述第二任务对应的调制信息的信息相似度，与所述任务网络中的第一网络节点和第二网络节点中网络节点的重复率正相关，其中，所述第一网络节点为所述任务网络中用于处理第一任务的待处理信息的目标网络节点，所述第二网络节点为所述任务网络中用于处理第二任务的待处理信息的目标网络节点。In a possible implementation, the preset task includes a first task and a second task, and the information similarity between the modulation information corresponding to the first task and the modulation information corresponding to the second task is positively correlated with the repetition rate of the network nodes in the first network node and the second network node in the task network, wherein the first network node is the target network node in the task network for processing the information to be processed of the first task, and the second network node is the target network node in the task network for processing the information to be processed of the second task.
在一种可能的实现方式中，如果第一任务和第二任务的任务相似度较高，则第一任务对应的调制信息与第二任务对应的调制信息的信息相似度也较高，进一步地，通过相似度较高的调制信息选择出的网络节点的重复率较高。即，在任务相似度较高的情况下，在执行任务或进行训练时，可服用一部分网络节点以及这些网络节点的参数，以提高训练效率和参数利用率。例如，第一任务和第二任务均为图像处理任务，第一任务为对图像中目标对象的位置进行识别的任务，第二任务为对图像中目标对象的轮廓进行识别的任务，第一任务和第二任务的相似度较高，则第一网络节点和第二网络节点的重复率较高，即，执行第一任务和第二任务使用的网络节点中，重复的网络节点的数量较多，可提升网络节点的参数利用率。In a possible implementation, if the task similarity between the first task and the second task is high, the information similarity between the modulation information corresponding to the first task and the modulation information corresponding to the second task is also high, and further, the repetition rate of the network nodes selected by the modulation information with high similarity is high. That is, when the task similarity is high, when executing tasks or training, a part of the network nodes and the parameters of these network nodes can be used to improve the training efficiency and parameter utilization. For example, the first task and the second task are both image processing tasks, the first task is a task for identifying the position of the target object in the image, and the second task is a task for identifying the outline of the target object in the image. The similarity between the first task and the second task is high, then the repetition rate of the first network node and the second network node is high, that is, among the network nodes used to execute the first task and the second task, the number of repeated network nodes is large, which can improve the parameter utilization of the network nodes.
又例如，第一任务为对图像中目标对象的位置进行识别的任务，第二任务为对文字的语义进行识别的任务，第一任务和第二任务的相似度较低，则第一网络节点和第二网络节点的重复率较低，即，执行第一任务和第二任务使用的网络节点中，重复的网络节点的数量较少，可减少不同任务之间的参数干扰。For another example, the first task is to identify the position of a target object in an image, and the second task is to identify the semantics of text. If the similarity between the first task and the second task is low, then the repetition rate of the first network node and the second network node is low, that is, among the network nodes used to execute the first task and the second task, the number of repeated network nodes is small, which can reduce parameter interference between different tasks.
在一种可能的实现方式中，在步骤S13中，可通过目标网络节点来处理待处理信息，获得目标任务的处理结果。例如，可通过第一网络节点来处理第一任务的待处理信息，并获得第一任务的处理结果，并可通过第二网络节点来处理第二任务的待处理信息，并获得第二任务的处理结果。即，可通过多个任务的调制信息选择与各任务对应的目标网络节点，并分别使用与各任务对应的目标网络节点处理各任务的待处理信息，进而获得各任务的处理结果。In a possible implementation, in step S13, the information to be processed may be processed by the target network node to obtain the processing result of the target task. For example, the information to be processed of the first task may be processed by the first network node to obtain the processing result of the first task, and the information to be processed of the second task may be processed by the second network node to obtain the processing result of the second task. That is, the target network node corresponding to each task may be selected by the modulation information of multiple tasks, and the target network node corresponding to each task may be used to process the information to be processed of each task, thereby obtaining the processing result of each task.
在示例中，任务A为图像处理任务，待处理信息为图像信息，任务B为文字处理任务，待处理信息为文字信息，任务C为语音处理任务，待处理信息为语音信息。调制网络可针对图像信息生成任务A的调制信息A，针对文字信息生成任务B的调制信息B，针对语音信息生成任务C的调制信息C。在执行任务时，任务网络可根据调制信息A确定用于执行任务A的网络节点组A，并通过网络节点组A来处理图像信息，获得任务A的处理结果，并可根据调制信息B确定用于执行任务B的网络节点组B，并通过网络节点组B来处理文字信息，获得任务B的处理结果，并可根据调制信息C确定用于执行任务C的网络节点组C，并通过网络节点组C来处理语音信息，获得任务C的处理结果。In the example, task A is an image processing task, and the information to be processed is image information; task B is a text processing task, and the information to be processed is text information; task C is a voice processing task, and the information to be processed is voice information. The modulation network can generate modulation information A for task A for image information, modulation information B for task B for text information, and modulation information C for task C for voice information. When executing tasks, the task network can determine the network node group A for executing task A according to the modulation information A, and process the image information through the network node group A to obtain the processing result of task A, and can determine the network node group B for executing task B according to the modulation information B, and process the text information through the network node group B to obtain the processing result of task B, and can determine the network node group C for executing task C according to the modulation information C, and process the voice information through the network node group C to obtain the processing result of task C.
在一种可能的实现方式中，可对任务网络和调制网络进行训练，使调制网络获得根据预设任务的待处理信息获得与预设任务对应的调制信息的能力，并使任务网络获得处理任务的能力。In a possible implementation, the task network and the modulation network may be trained so that the modulation network acquires the ability to obtain modulation information corresponding to a preset task based on information to be processed of the preset task, and the task network acquires the ability to process tasks.
在一种可能的实现方式中，可首先训练调制网络，并通过已训练的调制网络生成调制信息，以根据调制信息选择任务网络中用于执行特定任务的目标网络节点，进而利用该特定任务的样本信息来训练任务网络中的目标网络节点的网络参数。In one possible implementation, the modulation network can be trained first, and modulation information can be generated through the trained modulation network to select the target network node in the task network for performing a specific task according to the modulation information, and then the sample information of the specific task can be used to train the network parameters of the target network node in the task network.
在一种可能的实现方式中，训练所述调制网络的步骤可包括：将训练任务的训练样本输入所述调制网络，获得所述训练样本的第一训练调制信息，所述训练任务是多个任务中的任意一个；根据所述训练任务的多个训练样本的第一训练调制信息，获得所述训练任务的第二训练调制信息；根据所述第一训练调制信息和所述第二训练调制信息，训练所述调制网络。In a possible implementation, the step of training the modulation network may include: inputting a training sample of a training task into the modulation network to obtain first training modulation information of the training sample, where the training task is any one of multiple tasks; obtaining second training modulation information of the training task based on the first training modulation information of multiple training samples of the training task; and training the modulation network based on the first training modulation information and the second training modulation information.
在一种可能的实现方式中，所述训练任务可以是任意一种任务，例如，图像处理任务、语音处理任务、文字处理任务、向量处理任务等，训练任务的训练样本可以是图像信息、语音信息、文字信息、向量信息等，本公开对训练任务和训练样本不做限制。In one possible implementation, the training task can be any type of task, for example, an image processing task, a speech processing task, a text processing task, a vector processing task, etc. The training sample of the training task can be image information, speech information, text information, vector information, etc. The present disclosure does not limit the training tasks and training samples.
在一种可能的实现方式中，调制网络可根据输入的训练样本，输出第一训练调制信息，在示例中，第一训练调制信息可以是向量形式的输出信息，该向量可包括多个元素，可利用向量中的元素来确定任务网络中屏蔽的网络节点。本公开对第一训练调制信息的数据类型不做限制。In a possible implementation, the modulation network may output first training modulation information according to the input training sample. In an example, the first training modulation information may be output information in a vector form. The vector may include multiple elements. The elements in the vector may be used to determine the shielded network nodes in the task network. The present disclosure does not limit the data type of the first training modulation information.
在一种可能的实现方式中，每个训练任务可包括多个训练样本，可根据调制网络输出的该训练任务的多个训练样本的第一训练调制信息，来确定该训练任务的第二训练调制信息。例如，可对多个训练样本的第一训练调制信息进行加权平均，或者取各第一训练调制信息中对应元素的最大值等，本公开对根据第一训练调制信息获得训练任务的第二训练调制信息的方式不做限制。In a possible implementation, each training task may include multiple training samples, and the second training modulation information of the training task may be determined according to the first training modulation information of the multiple training samples of the training task output by the modulation network. For example, the first training modulation information of the multiple training samples may be weighted averaged, or the maximum value of the corresponding elements in each first training modulation information may be taken, etc. The present disclosure does not limit the method of obtaining the second training modulation information of the training task according to the first training modulation information.
在一种可能的实现方式中，根据所述训练任务的多个训练样本的第一训练调制信息，获得所述训练任务的第二训练调制信息，包括：获取所述多个训练样本的第一训练调制信息的平均训练调制信息；对所述平均训练调制信息进行阈值化处理，获得所述第二训练调制信息。In a possible implementation, obtaining second training modulation information of the training task based on first training modulation information of multiple training samples of the training task includes: obtaining average training modulation information of the first training modulation information of the multiple training samples; and performing thresholding processing on the average training modulation information to obtain the second training modulation information.
在一种可能的实现方式中，可对多个训练样本的第一训练调制信息进行加权平均处理，获得平均训练调制信息。在示例中，可针对多个第一训练调制信息中的对应元素进行加权平均，例如，针对每个第一训练调制信息中的第i(i为正整数)个元素进行加权平均(例如，每个第一训练调制信息的权值均相等)，获得平均训练调制信息中的第i个元素，可通过这种方式，确定平均训练调制信息中的各元素，即，获得平均训练调制信息。In a possible implementation, weighted averaging may be performed on the first training modulation information of the plurality of training samples to obtain average training modulation information. In the example, weighted averaging may be performed on corresponding elements in the plurality of first training modulation information, for example, weighted averaging may be performed on the i-th (i is a positive integer) element in each first training modulation information (for example, the weights of each first training modulation information are equal) element to obtain the i-th element in the average training modulation information. In this way, each element in the average training modulation information may be determined, that is, the average training modulation information may be obtained.
在一种可能的实现方式中，可将平均训练调制信息中的各元素进行阈值化处理，例如，可使各元素与阈值(例如，0)进行比较，如果该元素大于0，则保留该元素，如果该元素小于0，则将该元素替换为0，可通过该方式获得第二训练调制信息的各元素，即，获得第二训练调制信息。在另一示例中，可以1/2作为阈值，例如，可将平均训练调制信息的第i个元素减去1/2后，与0进行比较，如果第i个元素减去1/2后仍大于0，则可保留该元素，否则将该元素替换为0，可通过该方式获得第二训练调制信息的各元素，即，获得第二训练调制信息。本公开对进行比较的阈值不做限制。In one possible implementation, each element in the average training modulation information may be thresholded. For example, each element may be compared with a threshold value (e.g., 0). If the element is greater than 0, the element is retained. If the element is less than 0, the element is replaced with 0. In this way, each element of the second training modulation information is obtained, that is, the second training modulation information is obtained. In another example, 1/2 may be used as a threshold. For example, the ith element of the average training modulation information may be subtracted from 1/2 and then compared with 0. If the ith element is still greater than 0 after subtracting 1/2, the element may be retained. Otherwise, the element is replaced with 0. In this way, each element of the second training modulation information is obtained, that is, the second training modulation information is obtained. The present disclosure does not limit the threshold value for comparison.
在一种可能的实现方式中，可根据训练任务的第二训练调制信息，以及训练任务的多个训练样本的第一训练调制信息训练调制网络。例如，多个训练样本对应的任务均为同一个训练任务，则训练样本对应的第一训练调制信息应相同。在示例中，可对各第一训练调制信息施加相似度约束，例如，使调制网络的输出的各第一训练调制信息均与训练任务的第二训练调制信息相似，例如，可通过训练使各第一训练调制信息逼近第二训练调制信息，使得调制网络针对同一训练任务的多个训练样本输出的调制信息(即，第一训练调制信息)相同。In a possible implementation, the modulation network may be trained according to the second training modulation information of the training task and the first training modulation information of the multiple training samples of the training task. For example, if the tasks corresponding to the multiple training samples are the same training task, the first training modulation information corresponding to the training samples should be the same. In the example, a similarity constraint may be imposed on each first training modulation information, for example, so that each first training modulation information output by the modulation network is similar to the second training modulation information of the training task. For example, each first training modulation information may be trained to approximate the second training modulation information, so that the modulation information (i.e., the first training modulation information) output by the modulation network for multiple training samples of the same training task is the same.
在一种可能的实现方式中，根据所述第一训练调制信息和所述第二训练调制信息，训练所述调制网络，包括：根据所述第一训练调制信息和所述第二训练调制信息，获得相似度约束信息；根据所述第二训练调制信息，获得正则化信息；根据所述相似度约束信息和所述正则化信息，确定所述调制网络的第一网络损失；根据所述第一网络损失，训练所述调制网络。In a possible implementation, the modulation network is trained according to the first training modulation information and the second training modulation information, including: obtaining similarity constraint information according to the first training modulation information and the second training modulation information; obtaining regularization information according to the second training modulation information; determining a first network loss of the modulation network according to the similarity constraint information and the regularization information; and training the modulation network according to the first network loss.
在一种可能的实现方式中，根据所述第一训练调制信息和所述第二训练调制信息，获得相似度约束信息，即，通过使各第一训练调制信息均逼近训练任务的第二训练调制信息来约束各第一训练调制信息的相似度。在示例中，相似度约束信息可包括将第一训练调制信息与第二训练调制信息获得的向量的二范数，通过训练，使该二范数最小化，从而使得调制网络输出的各训练样本的第一训练调制信息逼近训练任务的第二训练调制信息。In a possible implementation, similarity constraint information is obtained based on the first training modulation information and the second training modulation information, that is, the similarity of each first training modulation information is constrained by making each first training modulation information approach the second training modulation information of the training task. In an example, the similarity constraint information may include a second norm of a vector obtained by combining the first training modulation information and the second training modulation information, and the second norm is minimized through training, so that the first training modulation information of each training sample output by the modulation network approaches the second training modulation information of the training task.
在一种可能的实现方式中，通过相似度约束信息，可使调制网络输出的相同或相似任务的调制信息的相似度较高，并且，并未约束调制网络输出不同任务的调制信息。使得调制网络输出相同任务的调制信息相似度较高，在特征空间中更加集中，并使得调制网络输出不同任务的调制信息相似度不高，在特征空间中相互远离。In a possible implementation, similarity constraint information can be used to make the modulation information of the same or similar tasks output by the modulation network have a high similarity, and the modulation information of different tasks output by the modulation network is not constrained. This makes the modulation information of the same task output by the modulation network have a high similarity and more concentrated in the feature space, and makes the modulation information of different tasks output by the modulation network have a low similarity and are far away from each other in the feature space.
在一种可能的实现方式中，可通过正则化信息，使得调制网络的网络损失为正值，使得调制网络可进行多次训练，避免发生梯度消失等状况。在示例中，可通过训练任务的第二训练调制信息的一范数来确定正则化信息，本公开对正则化信息的具体形式不做限制，只需使得调制网络的网络损失为正值即可。In a possible implementation, the network loss of the modulation network can be made positive through regularization information, so that the modulation network can be trained multiple times to avoid gradient disappearance and other conditions. In the example, the regularization information can be determined by a norm of the second training modulation information of the training task. The present disclosure does not limit the specific form of the regularization information, and it is only necessary to make the network loss of the modulation network positive.
在一种可能的实现方式中，可根据相似度约束信息和正则化信息，确定所述调制网络的第一网络损失，例如，可使相似度约束信息和正则化信息相加，获得第一网络损失。进一步地，可通过第一网络损失训练调制网络，例如，可将第一网络损失进行反向传播，并通过梯度下降法来调节调制网络的网络参数。可输入多个训练任务的多个训练样本，以迭代执行上述训练步骤，直到满足调制网络的训练条件，例如，训练次数达到预定次数、第一网络损失小于预设阈值或收敛于预设区间等，本公开对调制网络的训练条件不做限制。In one possible implementation, the first network loss of the modulation network can be determined based on the similarity constraint information and the regularization information. For example, the similarity constraint information and the regularization information can be added to obtain the first network loss. Furthermore, the modulation network can be trained by the first network loss. For example, the first network loss can be back-propagated, and the network parameters of the modulation network can be adjusted by the gradient descent method. Multiple training samples of multiple training tasks can be input to iteratively execute the above training steps until the training conditions of the modulation network are met, for example, the number of training times reaches a predetermined number of times, the first network loss is less than a preset threshold or converges to a preset interval, etc. The present disclosure does not limit the training conditions of the modulation network.
通过这种方式，可通过相似度约束信息，使调制网络输出相同任务的调制信息相似度较高，在特征空间中更加集中，并使调制网络输出不同任务的调制信息相似度不高，在特征空间中相互远离。使得任务网络通过调制网络输出的调制信息选择目标网络节点时，针对相似的任务选择的目标网络节点重复率较高，针对不相似的任务选择的目标网络节点的重复率较低，可提升网络参数的利用效率，提升任务网络的训练效率，并减少不同任务之间的参数干扰。In this way, similarity constraint information can be used to make the modulation information of the same task output by the modulation network have a high similarity and be more concentrated in the feature space, and the modulation information of different tasks output by the modulation network have a low similarity and are far away from each other in the feature space. When the task network selects the target network node through the modulation information output by the modulation network, the repetition rate of the target network nodes selected for similar tasks is high, and the repetition rate of the target network nodes selected for dissimilar tasks is low, which can improve the utilization efficiency of network parameters, improve the training efficiency of the task network, and reduce parameter interference between different tasks.
在一种可能的实现方式中，在调制网络训练完成后，可利用已训练的调制网络生成调制信息，来训练任务网络。例如，可利用调制信息选择目标网络节点，并调整目标网络节点的网络参数，可针对多个训练任务对任务网络进行训练，获得训练后的任务网络。训练所述任务网络的步骤可包括：根据训练任务的训练样本以及已训练的调制网络，获得所述训练任务的第三训练调制信息，所述训练任务是多个任务中的任意一个；根据所述第三训练调制信息确定所述任务网络中用于处理所述训练任务的训练样本的目标网络节点；通过所述目标网络节点处理所述训练任务的训练样本，获得训练结果；根据所述训练结果与所述训练样本的标注信息确定所述任务网络的第二网络损失；根据所述第二网络损失，训练所述任务网络。In a possible implementation, after the training of the modulation network is completed, the trained modulation network can be used to generate modulation information to train the task network. For example, the modulation information can be used to select the target network node and adjust the network parameters of the target network node. The task network can be trained for multiple training tasks to obtain the trained task network. The step of training the task network may include: obtaining the third training modulation information of the training task according to the training sample of the training task and the trained modulation network, the training task is any one of the multiple tasks; determining the target network node in the task network for processing the training sample of the training task according to the third training modulation information; processing the training sample of the training task through the target network node to obtain the training result; determining the second network loss of the task network according to the training result and the labeling information of the training sample; and training the task network according to the second network loss.
在一种可能的实现方式中，可将训练任务的训练样本输入已训练的调制网络，获得第三训练调制信息。在示例中，如果任务网络的网络层级较多，网络节点也较多，则选择目标网络节点的选择空间较大，不同任务之间发生参数干扰的可能性较小，可直接将已训练的调制网络输出的调制信息确定为第三训练调制信息。In a possible implementation, the training samples of the training task can be input into the trained modulation network to obtain the third training modulation information. In the example, if the task network has more network levels and more network nodes, the selection space for selecting the target network node is larger, and the possibility of parameter interference between different tasks is smaller, and the modulation information output by the trained modulation network can be directly determined as the third training modulation information.
在一种可能的实现方式中，如果任务网络的网络层级较少，例如，任务网络为单隐层神经网络，则选择目标网络节点的选择空间较小，不同任务之间发生参数干扰的可能性较大，可通过当前训练任务的调制信息以及历史训练任务的调制信息，确定第三训练调制信息，进而确定未训练过的网络节点，作为目标网络节点，以减少不同任务之间的参数干扰。In one possible implementation, if the task network has fewer network levels, for example, the task network is a single hidden layer neural network, then the selection space for selecting the target network node is smaller, and the possibility of parameter interference between different tasks is greater. The modulation information of the current training task and the modulation information of the historical training tasks can be used to determine the third training modulation information, and then determine the untrained network node as the target network node to reduce parameter interference between different tasks.
在一种可能的实现方式中，根据训练任务的训练样本以及已训练的调制网络，获得所述训练任务的第三训练调制信息，包括：在所述任务网络为单隐层神经网络的情况下，根据所述调制网络获得的当前训练任务的第四训练调制信息，以及已训练的历史训练任务的第五训练调制信息，确定当前训练任务的第三训练调制信息。In one possible implementation, third training modulation information of the training task is obtained based on the training samples of the training task and the trained modulation network, including: when the task network is a single hidden layer neural network, the third training modulation information of the current training task is determined based on the fourth training modulation information of the current training task obtained from the modulation network and the fifth training modulation information of the trained historical training task.
在示例中，如果当前训练任务为第一个训练任务，则不存在已训练的历史训练任务，可通过调制网络输出的调制信息选择目标网络节点，不存在不同任务之间的参数干扰。如果当前训练任务不是第一个训练任务，则可通过以下公式(1)确定第三训练调制信息，以选择与历史训练任务训练过的网络节点不同的网络节点(即，未训练过的网络节点)进行训练，从而减少不同任务之间的参数干扰。In the example, if the current training task is the first training task, there is no historical training task that has been trained, and the target network node can be selected by the modulation information output by the modulation network, and there is no parameter interference between different tasks. If the current training task is not the first training task, the third training modulation information can be determined by the following formula (1) to select a network node different from the network node trained by the historical training task (i.e., an untrained network node) for training, thereby reducing parameter interference between different tasks.
其中，maskk为调制网络针对当前训练任务(第k个训练任务)输出的第四训练调制信息，maski为调制网络针对历史训练任务(第i个训练任务)输出的第五训练调制信息，mask′k为所述第三训练调制信息。Among them, maskk is the fourth training modulation information output by the modulation network for the current training task (kth training task), maski is the fifth training modulation information output by the modulation network for the historical training task (ith training task), and mask′k is the third training modulation information.
在一种可能的实现方式中，任务网络可通过第三训练调制信息屏蔽当前训练任务中不进行训练的网络节点，并选择进行训练的目标网络节点。进一步地，可通过目标网络节点来处理训练样本，获得训练结果，以根据训练结果与训练样本的标注信息之间的误差确定任务网络的第二网络损失，并通过第二网络损失来训练任务网络。In a possible implementation, the task network can shield the network nodes that are not to be trained in the current training task through the third training modulation information, and select the target network nodes to be trained. Further, the training samples can be processed through the target network nodes to obtain the training results, so as to determine the second network loss of the task network according to the error between the training results and the labeled information of the training samples, and train the task network through the second network loss.
在一种可能的实现方式中，可将第二网络损失进行反向传播，并通过梯度下降法来调节任务网络的目标网络节点的网络参数。可输入多个训练任务的多个训练样本，以迭代执行上述训练步骤，直到满足任务网络的训练条件，例如，训练次数达到预定次数、第二网络损失小于预设阈值或收敛于预设区间，任务网络中所有网络节点均已经过预定次数的训练等，本公开对任务网络的训练条件不做限制。In a possible implementation, the second network loss can be back-propagated, and the network parameters of the target network nodes of the task network can be adjusted by the gradient descent method. Multiple training samples of multiple training tasks can be input to iteratively perform the above training steps until the training conditions of the task network are met, for example, the number of trainings reaches a predetermined number of times, the second network loss is less than a preset threshold or converges to a preset interval, all network nodes in the task network have been trained a predetermined number of times, etc. The present disclosure does not limit the training conditions of the task network.
在一种可能的实现方式中，在调制网络与任务网络均训练完成后，可对上述两个网络进行测试，如果在测试中执行任务的准确率高于准确率阈值，则可将调制网络与任务网络用于实际执行任务。否则，可继续训练上述两个网络。In a possible implementation, after the modulation network and the task network are trained, the two networks can be tested. If the accuracy of executing the task in the test is higher than the accuracy threshold, the modulation network and the task network can be used to actually execute the task. Otherwise, the two networks can continue to be trained.
根据本公开的实施例的任务处理方法，可通过调制网络生成的调制信息对任务网络中用于处理任务的网络节点进行选择，可减少参数间的相互干扰，缓解灾难性遗忘，并通过调制信息可提高相似任务中网络节点的复用率，提高网络参数的利用效率。并且，可通过脉冲神经网络处理作为任务网络，处理复杂的时序任务，可使脉冲神经网络在调制信号的作用下进行连续多个任务的训练或执行多个任务，提高了神经网络的训练和执行效率。进一步地，在训练调制网络的过程中，可通过相似度约束信息，使调制网络输出相同任务的调制信息相似度较高，在特征空间中更加集中，并使调制网络输出不同任务的调制信息相似度不高，在特征空间中相互远离。使得任务网络通过调制网络输出的调制信息选择目标网络节点时，针对相似的任务选择的目标网络节点重复率较高，针对不相似的任务选择的目标网络节点的重复率较低，可提升网络参数的利用效率，提升任务网络的训练效率，并减少不同任务之间的参数干扰。According to the task processing method of the embodiment of the present disclosure, the modulation information generated by the modulation network can be used to select the network nodes used to process the task in the task network, which can reduce the mutual interference between parameters, alleviate catastrophic forgetting, and improve the reuse rate of network nodes in similar tasks through the modulation information, thereby improving the utilization efficiency of network parameters. In addition, the pulse neural network can be used as a task network to process complex sequential tasks, so that the pulse neural network can be trained or execute multiple tasks continuously under the action of the modulation signal, thereby improving the training and execution efficiency of the neural network. Further, in the process of training the modulation network, the modulation information output by the modulation network for the same task can be made to have a higher similarity and be more concentrated in the feature space through the similarity constraint information, and the modulation information output by the modulation network for different tasks can be made to have a lower similarity and be far away from each other in the feature space. When the task network selects the target network node through the modulation information output by the modulation network, the repetition rate of the target network node selected for similar tasks is higher, and the repetition rate of the target network node selected for dissimilar tasks is lower, thereby improving the utilization efficiency of network parameters, improving the training efficiency of the task network, and reducing the parameter interference between different tasks.
图2示出根据本公开实施例的任务处理方法的应用示意图，如图2所示，任务网络可通过调制信息选择不同的节点，来执行三个任务(即，任务1、任务2和任务3)。FIG2 shows an application diagram of the task processing method according to an embodiment of the present disclosure. As shown in FIG2 , the task network can select different nodes by modulating information to execute three tasks (ie, task 1, task 2, and task 3).
在一种可能的实现方式中，调制网络可针对各任务的待处理信息，输出各任务的调制信息，例如，针对任务1的待处理信息，可输出任务1的调制信息1，针对任务2的待处理信息，可输出任务2的调制信息2，针对任务3的待处理信息，可输出任务3的调制信息3。In one possible implementation, the modulation network can output modulation information of each task with respect to the information to be processed of each task. For example, with respect to the information to be processed of task 1, modulation information 1 of task 1 can be output; with respect to the information to be processed of task 2, modulation information 2 of task 2 can be output; and with respect to the information to be processed of task 3, modulation information 3 of task 3 can be output.
在一种可能的实现方式中，任务网络可根据调制信息，屏蔽部分网络节点，即，被屏蔽的网络节点不参与当前任务的执行，未被屏蔽的网络节点即为用于执行当前任务的目标网络节点。在示例中，如图2所示，在执行任务2时，可屏蔽第一个隐层中的第1个和第4个网络节点，以及第二个隐层中的第2个和第3个网络节点。即，第一个隐层中的第2个和第3个网络节点以及第二个隐层中的第1个和第4个网络节点为执行任务2的目标网络节点。In a possible implementation, the task network can shield some network nodes according to the modulation information, that is, the shielded network nodes do not participate in the execution of the current task, and the unshielded network nodes are the target network nodes for executing the current task. In the example, as shown in FIG2, when executing Task 2, the 1st and 4th network nodes in the first hidden layer and the 2nd and 3rd network nodes in the second hidden layer can be shielded. That is, the 2nd and 3rd network nodes in the first hidden layer and the 1st and 4th network nodes in the second hidden layer are the target network nodes for executing Task 2.
在一种可能的实现方式中，任务网络可通过目标网络节点处理任务2的待处理信息，获得任务2的处理结果。类似地，任务网络可通过调制信息1选择任务1的目标网络节点，并通过目标网络节点处理任务1的待处理信息，获得任务1的处理结果。还可通过调制信息3选择任务3的目标网络节点，并通过目标网络节点处理任务3的待处理信息，获得任务3的处理结果。In a possible implementation, the task network can process the information to be processed of task 2 through the target network node to obtain the processing result of task 2. Similarly, the task network can select the target network node of task 1 through modulation information 1, and process the information to be processed of task 1 through the target network node to obtain the processing result of task 1. The target network node of task 3 can also be selected through modulation information 3, and the information to be processed of task 3 can be processed through the target network node to obtain the processing result of task 3.
在一种可能的实现方式中，所述任务处理方法可用于使神经网络处理多种任务的场景，例如，在监控领域中，可确定目标对象的位置，并识别目标对象的身份。在智能家居领域中，可识别使用者的身份，并识别使用者的语音指令等。本公开对任务处理方法的应用领域不做限制。In a possible implementation, the task processing method can be used to enable a neural network to process multiple tasks. For example, in the field of monitoring, the location of a target object can be determined and the identity of the target object can be identified. In the field of smart homes, the identity of a user can be identified and the user's voice commands can be recognized. The present disclosure does not limit the application field of the task processing method.
图3示出根据本公开实施例的任务处理装置的框图，如图3所示，所述装置包括：调制信息模块11，用于将目标任务的待处理信息输入调制网络，获得与所述目标任务对应的调制信息，所述目标任务是多个预设任务中的任意一个；目标节点模块12，用于根据所述调制信息，从任务网络的多个网络节点中，确定出用于处理所述待处理信息的目标网络节点；处理模块13，用于通过所述目标网络节点对所述待处理信息进行处理，获得所述目标任务的处理结果。Figure 3 shows a block diagram of a task processing device according to an embodiment of the present disclosure. As shown in Figure 3, the device includes: a modulation information module 11, which is used to input the information to be processed of the target task into the modulation network to obtain modulation information corresponding to the target task, and the target task is any one of a plurality of preset tasks; a target node module 12, which is used to determine, according to the modulation information, a target network node for processing the information to be processed from a plurality of network nodes of the task network; and a processing module 13, which is used to process the information to be processed through the target network node to obtain a processing result of the target task.
在一种可能的实现方式中，所述预设任务包括第一任务和第二任务，所述第一任务对应的调制信息与所述第二任务对应的调制信息的信息相似度，与所述第一任务和所述第二任务的任务相似度正相关。In a possible implementation, the preset task includes a first task and a second task, and information similarity between modulation information corresponding to the first task and modulation information corresponding to the second task is positively correlated with task similarity between the first task and the second task.
在一种可能的实现方式中，所述预设任务包括第一任务和第二任务，所述第一任务对应的调制信息与所述第二任务对应的调制信息的信息相似度，与所述任务网络中的第一网络节点和第二网络节点中网络节点的重复率正相关，其中，所述第一网络节点为所述任务网络中用于处理第一任务的待处理信息的目标网络节点，所述第二网络节点为所述任务网络中用于处理第二任务的待处理信息的目标网络节点。In a possible implementation, the preset task includes a first task and a second task, and the information similarity between the modulation information corresponding to the first task and the modulation information corresponding to the second task is positively correlated with the repetition rate of the network nodes in the first network node and the second network node in the task network, wherein the first network node is the target network node in the task network for processing the information to be processed of the first task, and the second network node is the target network node in the task network for processing the information to be processed of the second task.
在一种可能的实现方式中，所述调制网络包括人工神经网络和脉冲神经网络中的任意一种，所述任务网络包括脉冲神经网络和人工神经网络中的任意一种。在一种可能的实现方式中，所述装置还包括调制网络训练模块，所述调制网络训练模块用于：将训练任务的训练样本输入所述调制网络，获得所述训练样本的第一训练调制信息，所述训练任务是多个任务中的任意一个；根据所述训练任务的多个训练样本的第一训练调制信息，获得所述训练任务的第二训练调制信息；根据所述第一训练调制信息和所述第二训练调制信息，训练所述调制网络。In a possible implementation, the modulation network includes any one of an artificial neural network and a pulse neural network, and the task network includes any one of a pulse neural network and an artificial neural network. In a possible implementation, the device further includes a modulation network training module, and the modulation network training module is used to: input a training sample of a training task into the modulation network to obtain first training modulation information of the training sample, and the training task is any one of a plurality of tasks; obtain second training modulation information of the training task according to the first training modulation information of the plurality of training samples of the training task; and train the modulation network according to the first training modulation information and the second training modulation information.
在一种可能的实现方式中，所述调制网络训练模块进一步用于：获取所述多个训练样本的第一训练调制信息的平均训练调制信息；对所述平均训练调制信息进行阈值化处理，获得所述第二训练调制信息。In a possible implementation, the modulation network training module is further used to: obtain average training modulation information of first training modulation information of the multiple training samples; and perform thresholding processing on the average training modulation information to obtain the second training modulation information.
在一种可能的实现方式中，所述调制网络训练模块进一步用于：根据所述第一训练调制信息和所述第二训练调制信息，获得相似度约束信息；根据所述第二训练调制信息，获得正则化信息；根据所述相似度约束信息和所述正则化信息，确定所述调制网络的第一网络损失；根据所述第一网络损失，训练所述调制网络。In a possible implementation, the modulation network training module is further used to: obtain similarity constraint information based on the first training modulation information and the second training modulation information; obtain regularization information based on the second training modulation information; determine the first network loss of the modulation network based on the similarity constraint information and the regularization information; and train the modulation network based on the first network loss.
在一种可能的实现方式中，所述装置还包括任务网络训练模块，所述任务网络训练模块用于：根据训练任务的训练样本以及已训练的调制网络，获得所述训练任务的第三训练调制信息，所述训练任务是多个任务中的任意一个；根据所述第三训练调制信息确定所述任务网络中用于处理所述训练任务的训练样本的目标网络节点；通过所述目标网络节点处理所述训练任务的训练样本，获得训练结果；根据所述训练结果与所述训练样本的标注信息确定所述任务网络的第二网络损失；根据所述第二网络损失，训练所述任务网络。In a possible implementation, the device also includes a task network training module, which is used to: obtain third training modulation information of the training task based on the training samples of the training task and a trained modulation network, where the training task is any one of multiple tasks; determine a target network node in the task network for processing the training samples of the training task based on the third training modulation information; process the training samples of the training task through the target network node to obtain a training result; determine a second network loss of the task network based on the training result and the labeling information of the training samples; and train the task network based on the second network loss.
在一种可能的实现方式中，所述任务网络训练模块进一步用于：在所述任务网络为单隐层神经网络的情况下，根据所述调制网络获得的当前训练任务的第四训练调制信息，以及已训练的历史训练任务的第五训练调制信息，确定当前训练任务的第三训练调制信息。In one possible implementation, the task network training module is further used to: when the task network is a single hidden layer neural network, determine the third training modulation information of the current training task based on the fourth training modulation information of the current training task obtained by the modulation network and the fifth training modulation information of the trained historical training tasks.
在一种可能的实现方式中，所述多个预设任务包括图像处理任务、语音处理任务、文字处理任务、向量处理任务中的至少一种，所述待处理信息包括图像信息、语音信息、文字信息、向量信息中的至少一种。In a possible implementation, the multiple preset tasks include at least one of an image processing task, a voice processing task, a text processing task, and a vector processing task, and the information to be processed includes at least one of image information, voice information, text information, and vector information.
可以理解，本公开提及的上述各个方法实施例，在不违背原理逻辑的情况下，均可以彼此相互结合形成结合后的实施例，限于篇幅，本公开不再赘述。It can be understood that the above-mentioned various method embodiments mentioned in the present disclosure can be combined with each other to form a combined embodiment without violating the principle logic. Due to space limitations, the present disclosure will not go into details.
此外，本公开还提供了任务处理装置、电子设备、计算机可读存储介质、程序，上述均可用来实现本公开提供的任一种任务处理方法，相应技术方案和描述和参见方法部分的相应记载，不再赘述。In addition, the present disclosure also provides a task processing device, an electronic device, a computer-readable storage medium, and a program, all of which can be used to implement any task processing method provided by the present disclosure. The corresponding technical solutions and descriptions are referred to the corresponding records in the method part and will not be repeated here.
本领域技术人员可以理解，在具体实施方式的上述方法中，各步骤的撰写顺序并不意味着严格的执行顺序而对实施过程构成任何限定，各步骤的具体执行顺序应当以其功能和可能的内在逻辑确定。Those skilled in the art will appreciate that, in the above method of specific implementation, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of the steps should be determined by their functions and possible internal logic.
在一些实施例中，本公开实施例提供的装置具有的功能或包含的模块可以用于执行上文方法实施例描述的方法，其具体实现可以参照上文方法实施例的描述，为了简洁，这里不再赘述In some embodiments, the functions or modules included in the apparatus provided in the embodiments of the present disclosure may be used to execute the method described in the above method embodiment. The specific implementation thereof may refer to the description of the above method embodiment. For the sake of brevity, it will not be repeated here.
本公开实施例还提出一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述方法。计算机可读存储介质可以是非易失性计算机可读存储介质。The embodiment of the present disclosure also provides a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the above method is implemented. The computer-readable storage medium may be a non-volatile computer-readable storage medium.
本公开实施例还提出一种电子设备，包括：处理器；用于存储处理器可执行指令的存储器；其中，所述处理器被配置为上述方法。The embodiment of the present disclosure further proposes an electronic device, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured as the above method.
在示例中，存储器中可存储指令(例如，调用数据或信息的指令，或者神经网络的处理指令)，还可存储多个任务的待处理信息，处理器在执行所述任务处理方法时，可调用指令来执行任务处理方法的步骤。例如，可调用指令，以执行将待处理信息输入调制网络的指令，并可调用调制网络的处理指令，以通过调制网络来获得调制信息。例如，可调用指令，以执行通过调制信息确定任务网络中的目标网络节点的指令，又例如，可调用任务网络的处理指令，以通过任务网络的目标网络节点对待处理信息进行处理，以获得处理结果。In the example, the memory may store instructions (e.g., instructions for calling data or information, or processing instructions for a neural network), and may also store information to be processed for multiple tasks. When the processor executes the task processing method, the instructions may be called to execute the steps of the task processing method. For example, an instruction may be called to execute an instruction for inputting the information to be processed into a modulation network, and a processing instruction of the modulation network may be called to obtain modulation information through the modulation network. For example, an instruction may be called to execute an instruction for determining a target network node in a task network through modulation information, and for another example, a processing instruction of a task network may be called to process the information to be processed through the target network node of the task network to obtain a processing result.
电子设备可以被提供为终端、服务器或其它形态的设备。The electronic device may be provided as a terminal, a server, or a device in other forms.
图4是根据一示例性实施例示出的一种任务处理装置800的框图。例如，电子设备800可以是移动电话，计算机，数字广播终端，消息收发设备，游戏控制台，平板设备，医疗设备，健身设备，个人数字助理等终端。Fig. 4 is a block diagram of a task processing device 800 according to an exemplary embodiment. For example, the electronic device 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, or the like.
参照图4，电子设备800可以包括以下一个或多个组件：处理组件802，存储器804，电源组件806，多媒体组件808，音频组件810，输入/输出(I/O)的接口812，传感器组件814，以及通信组件816。4 , the electronic device 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power component 806 , a multimedia component 808 , an audio component 810 , an input/output (I/O) interface 812 , a sensor component 814 , and a communication component 816 .
处理组件802通常控制电子设备800的整体操作，诸如与显示，电话呼叫，数据通信，相机操作和记录操作相关联的操作。处理组件802可以包括一个或多个处理器820来执行指令，以完成上述的方法的全部或部分步骤。此外，处理组件802可以包括一个或多个模块，便于处理组件802和其他组件之间的交互。例如，处理组件802可以包括多媒体模块，以方便多媒体组件808和处理组件802之间的交互。The processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.
存储器804被配置为存储各种类型的数据以支持在电子设备800的操作。这些数据的示例包括用于在电子设备800上操作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图片，视频等。存储器804可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，如静态随机存取存储器(SRAM)，电可擦除可编程只读存储器(EEPROM)，可擦除可编程只读存储器(EPROM)，可编程只读存储器(PROM)，只读存储器(ROM)，磁存储器，快闪存储器，磁盘或光盘。The memory 804 is configured to store various types of data to support operations on the electronic device 800. Examples of such data include instructions for any application or method operating on the electronic device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.
电源组件806为电子设备800的各种组件提供电力。电源组件806可以包括电源管理系统，一个或多个电源，及其他与为电子设备800生成、管理和分配电力相关联的组件。The power supply component 806 provides power to the various components of the electronic device 800. The power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 800.
多媒体组件808包括在所述电子设备800和用户之间的提供一个输出接口的屏幕。在一些实施例中，屏幕可以包括液晶显示器(LCD)和触摸面板(TP)。如果屏幕包括触摸面板，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括一个或多个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界，而且还检测与所述触摸或滑动操作相关的持续时间和压力。在一些实施例中，多媒体组件808包括一个前置摄像头和/或后置摄像头。当电子设备800处于操作模式，如拍摄模式或视频模式时，前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。The multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and/or a rear camera. When the electronic device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and/or the rear camera may receive external multimedia data. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.
音频组件810被配置为输出和/或输入音频信号。例如，音频组件810包括一个麦克风(MIC)，当电子设备800处于操作模式，如呼叫模式、记录模式和语音识别模式时，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器804或经由通信组件816发送。在一些实施例中，音频组件810还包括一个扬声器，用于输出音频信号。The audio component 810 is configured to output and/or input audio signals. For example, the audio component 810 includes a microphone (MIC), and when the electronic device 800 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 804 or sent via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.
I/O接口812为处理组件802和外围接口模块之间提供接口，上述外围接口模块可以是键盘，点击轮，按钮等。这些按钮可包括但不限于：主页按钮、音量按钮、启动按钮和锁定按钮。I/O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: a home button, a volume button, a start button, and a lock button.
传感器组件814包括一个或多个传感器，用于为电子设备800提供各个方面的状态评估。例如，传感器组件814可以检测到电子设备800的打开/关闭状态，组件的相对定位，例如所述组件为电子设备800的显示器和小键盘，传感器组件814还可以检测电子设备800或电子设备800一个组件的位置改变，用户与电子设备800接触的存在或不存在，电子设备800方位或加速/减速和电子设备800的温度变化。传感器组件814可以包括接近传感器，被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件814还可以包括光传感器，如CMOS或CCD图像传感器，用于在成像应用中使用。在一些实施例中，该传感器组件814还可以包括加速度传感器，陀螺仪传感器，磁传感器，压力传感器或温度传感器。The sensor assembly 814 includes one or more sensors for providing various aspects of status assessment for the electronic device 800. For example, the sensor assembly 814 can detect the open/closed state of the electronic device 800, the relative positioning of the components, such as the display and keypad of the electronic device 800, and the sensor assembly 814 can also detect the position change of the electronic device 800 or a component of the electronic device 800, the presence or absence of contact between the user and the electronic device 800, the orientation or acceleration/deceleration of the electronic device 800, and the temperature change of the electronic device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 814 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
通信组件816被配置为便于电子设备800和其他设备之间有线或无线方式的通信。电子设备800可以接入基于通信标准的无线网络，如WiFi，2G或3G，或它们的组合。在一个示例性实施例中，通信组件816经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中，所述通信组件816还包括近场通信(NFC)模块，以促进短程通信。例如，在NFC模块可基于射频识别(RFID)技术，红外数据协会(IrDA)技术，超宽带(UWB)技术，蓝牙(BT)技术和其他技术来实现。The communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices. The electronic device 800 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.
在示例性实施例中，电子设备800可以被一个或多个应用专用集成电路(ASIC)、数字信号处理器(DSP)、数字信号处理设备(DSPD)、可编程逻辑器件(PLD)、现场可编程门阵列(FPGA)、控制器、微控制器、微处理器或其他电子元件实现，用于执行上述方法。In an exemplary embodiment, the electronic device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above methods.
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器804，上述计算机程序指令可由电子设备800的处理器820执行以完成上述方法。In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 804 including computer program instructions, which can be executed by a processor 820 of an electronic device 800 to perform the above method.
图5是根据一示例性实施例示出的一种任务处理装置1900的框图。例如，电子设备1900可以被提供为一服务器。参照图5，电子设备1900包括处理组件1922，其进一步包括一个或多个处理器，以及由存储器1932所代表的存储器资源，用于存储可由处理组件1922的执行的指令，例如应用程序。存储器1932中存储的应用程序可以包括一个或一个以上的每一个对应于一组指令的模块。此外，处理组件1922被配置为执行指令，以执行上述方法。FIG5 is a block diagram of a task processing device 1900 according to an exemplary embodiment. For example, the electronic device 1900 may be provided as a server. Referring to FIG5 , the electronic device 1900 includes a processing component 1922, which further includes one or more processors, and a memory resource represented by a memory 1932 for storing instructions executable by the processing component 1922, such as an application. The application stored in the memory 1932 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 1922 is configured to execute instructions to perform the above method.
电子设备1900还可以包括一个电源组件1926被配置为执行电子设备1900的电源管理，一个有线或无线网络接口1950被配置为将电子设备1900连接到网络，和一个输入输出(I/O)接口1958。电子设备1900可以操作基于存储在存储器1932的操作系统，例如Windows ServerTM，Mac OS XTM，UnixTM,LinuxTM，FreeBSDTM或类似。The electronic device 1900 may also include a power supply component 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network, and an input/output (I/O) interface 1958. The electronic device 1900 may operate based on an operating system stored in the memory 1932, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™ or the like.
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器1932，上述计算机程序指令可由电子设备1900的处理组件1922执行以完成上述方法。In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 1932 including computer program instructions, which can be executed by the processing component 1922 of the electronic device 1900 to perform the above method.
本公开可以是系统、方法和/或计算机程序产品。计算机程序产品可以包括计算机可读存储介质，其上载有用于使处理器实现本公开的各个方面的计算机可读程序指令。The present disclosure may be a system, a method and/or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.
计算机可读存储介质可以是可以保持和存储由指令执行设备使用的指令的有形设备。计算机可读存储介质例如可以是――但不限于――电存储设备、磁存储设备、光存储设备、电磁存储设备、半导体存储设备或者上述的任意合适的组合。计算机可读存储介质的更具体的例子(非穷举的列表)包括：便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、静态随机存取存储器(SRAM)、便携式压缩盘只读存储器(CD-ROM)、数字多功能盘(DVD)、记忆棒、软盘、机械编码设备、例如其上存储有指令的打孔卡或凹槽内凸起结构、以及上述的任意合适的组合。这里的计算机可读存储介质不被解释为瞬时信号本身，诸如无线电波或者其他自由传播的电磁波、通过波导或其他传输媒介传播的电磁波(例如，通过光纤电缆的光脉冲)、或者通过电线传输的电信号。A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the above. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the above. The computer-readable storage medium here is not to be interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.
这里所描述的计算机可读程序指令可以从计算机可读存储介质下载到各个计算/处理设备，或者通过网络、例如因特网、局域网、广域网和/或无线网下载到外部计算机或外部存储设备。网络可以包括铜传输电缆、光纤传输、无线传输、路由器、防火墙、交换机、网关计算机和/或边缘服务器。每个计算/处理设备中的网络适配卡或者网络接口从网络接收计算机可读程序指令，并转发该计算机可读程序指令，以供存储在各个计算/处理设备中的计算机可读存储介质中。The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing/processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and/or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and/or edge servers. The network adapter card or network interface in each computing/processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing/processing device.
用于执行本公开操作的计算机程序指令可以是汇编指令、指令集架构(ISA)指令、机器指令、机器相关指令、微代码、固件指令、状态设置数据、或者以一种或多种编程语言的任意组合编写的源代码或目标代码，所述编程语言包括面向对象的编程语言—诸如Smalltalk、C++等，以及常规的过程式编程语言—诸如“C”语言或类似的编程语言。计算机可读程序指令可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络—包括局域网(LAN)或广域网(WAN)—连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。在一些实施例中，通过利用计算机可读程序指令的状态信息来个性化定制电子电路，例如可编程逻辑电路、现场可编程门阵列(FPGA)或可编程逻辑阵列(PLA)，该电子电路可以执行计算机可读程序指令，从而实现本公开的各个方面。The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be personalized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.
这里参照根据本公开实施例的方法、装置(系统)和计算机程序产品的流程图和/或框图描述了本公开的各个方面。应当理解，流程图和/或框图的每个方框以及流程图和/或框图中各方框的组合，都可以由计算机可读程序指令实现。Various aspects of the present disclosure are described herein with reference to the flowcharts and/or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each box in the flowchart and/or block diagram and the combination of each box in the flowchart and/or block diagram can be implemented by computer-readable program instructions.
这些计算机可读程序指令可以提供给通用计算机、专用计算机或其它可编程数据处理装置的处理器，从而生产出一种机器，使得这些指令在通过计算机或其它可编程数据处理装置的处理器执行时，产生了实现流程图和/或框图中的一个或多个方框中规定的功能/动作的装置。也可以把这些计算机可读程序指令存储在计算机可读存储介质中，这些指令使得计算机、可编程数据处理装置和/或其他设备以特定方式工作，从而，存储有指令的计算机可读介质则包括一个制造品，其包括实现流程图和/或框图中的一个或多个方框中规定的功能/动作的各个方面的指令。These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions/actions specified in one or more boxes in the flowchart and/or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and/or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions/actions specified in one or more boxes in the flowchart and/or block diagram.
也可以把计算机可读程序指令加载到计算机、其它可编程数据处理装置、或其它设备上，使得在计算机、其它可编程数据处理装置或其它设备上执行一系列操作步骤，以产生计算机实现的过程，从而使得在计算机、其它可编程数据处理装置、或其它设备上执行的指令实现流程图和/或框图中的一个或多个方框中规定的功能/动作。Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions/actions specified in one or more boxes in the flowchart and/or block diagram.
附图中的流程图和框图显示了根据本公开的多个实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段或指令的一部分，所述模块、程序段或指令的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个连续的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或动作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and/or flow chart, and the combination of the square boxes in the block diagram and/or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.
以上已经描述了本公开的各实施例，上述说明是示例性的，并非穷尽性的，并且也不限于所披露的各实施例。在不偏离所说明的各实施例的范围和精神的情况下，对于本技术领域的普通技术人员来说许多修改和变更都是显而易见的。本文中所用术语的选择，旨在最好地解释各实施例的原理、实际应用或对市场中的技术的改进，或者使本技术领域的其它普通技术人员能理解本文披露的各实施例。The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011261159.0ACN112328398B (en) | 2020-11-12 | 2020-11-12 | Task processing method and device, electronic device and storage medium |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011261159.0ACN112328398B (en) | 2020-11-12 | 2020-11-12 | Task processing method and device, electronic device and storage medium |
| Publication Number | Publication Date |
|---|---|
| CN112328398A CN112328398A (en) | 2021-02-05 |
| CN112328398Btrue CN112328398B (en) | 2024-09-27 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011261159.0AActiveCN112328398B (en) | 2020-11-12 | 2020-11-12 | Task processing method and device, electronic device and storage medium |
| Country | Link |
|---|---|
| CN (1) | CN112328398B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113255905B (en)* | 2021-07-16 | 2021-11-02 | 成都时识科技有限公司 | Signal processing method of neurons in impulse neural network and network training method |
| US12182702B2 (en)* | 2021-09-22 | 2024-12-31 | KDDI Research, Inc. | Method and information processing apparatus that perform transfer learning while suppressing occurrence of catastrophic forgetting |
| CN116938658A (en)* | 2022-03-31 | 2023-10-24 | 华为技术有限公司 | Data processing methods, devices and systems |
| CN114781633B (en)* | 2022-06-17 | 2022-10-14 | 电子科技大学 | A processor integrating artificial neural network and spiking neural network |
| CN116304676B (en)* | 2023-02-01 | 2023-11-17 | 瑞鞍星医疗科技(苏州)有限公司 | Brain activity state classification model training method, device and equipment |
| CN116663631B (en)* | 2023-07-28 | 2024-05-14 | 之江实验室 | A method, device, storage medium and electronic device for executing a task |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109919296A (en)* | 2017-12-12 | 2019-06-21 | 杭州海康威视数字技术股份有限公司 | A deep neural network training method, device and computer equipment |
| CN111368923A (en)* | 2020-03-05 | 2020-07-03 | 上海商汤智能科技有限公司 | Neural network training method and device, electronic equipment and storage medium |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10049307B2 (en)* | 2016-04-04 | 2018-08-14 | International Business Machines Corporation | Visual object recognition |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109919296A (en)* | 2017-12-12 | 2019-06-21 | 杭州海康威视数字技术股份有限公司 | A deep neural network training method, device and computer equipment |
| CN111368923A (en)* | 2020-03-05 | 2020-07-03 | 上海商汤智能科技有限公司 | Neural network training method and device, electronic equipment and storage medium |
| Publication number | Publication date |
|---|---|
| CN112328398A (en) | 2021-02-05 |
| Publication | Publication Date | Title |
|---|---|---|
| CN112328398B (en) | Task processing method and device, electronic device and storage medium | |
| KR102365890B1 (en) | Method, apparatus and storage medium for training of neural network | |
| US20210241117A1 (en) | Method for processing batch-normalized data, electronic device and storage medium | |
| CN110909815B (en) | Neural network training method, neural network training device, neural network processing device, neural network training device, image processing device and electronic equipment | |
| CN111612070B (en) | Image description generation method and device based on scene graph | |
| CN111259967B (en) | Image classification and neural network training method, device, equipment and storage medium | |
| CN113065591B (en) | Target detection method and device, electronic equipment and storage medium | |
| CN111931844B (en) | Image processing method and device, electronic equipment and storage medium | |
| CN111753917B (en) | Data processing method, device and storage medium | |
| WO2022011892A1 (en) | Network training method and apparatus, target detection method and apparatus, and electronic device | |
| CN110889469A (en) | Image processing method and device, electronic equipment and storage medium | |
| CN109165738B (en) | Neural network model optimization method and device, electronic device and storage medium | |
| CN109858614B (en) | Neural network training method and device, electronic equipment and storage medium | |
| CN111242303B (en) | Network training method and device, and image processing method and device | |
| CN110659690B (en) | Neural network construction method and device, electronic equipment and storage medium | |
| CN113298091A (en) | Image processing method and device, electronic equipment and storage medium | |
| CN113361540A (en) | Image processing method and device, electronic equipment and storage medium | |
| CN110188871A (en) | Computing method, device and related products | |
| CN109685041B (en) | Image analysis method and device, electronic equipment and storage medium | |
| CN112559673A (en) | Language processing model training method and device, electronic equipment and storage medium | |
| CN115495364A (en) | System fault diagnosis method, apparatus, storage medium, and program product | |
| CN116543211A (en) | Image attribute editing method, device, electronic device and storage medium | |
| CN110502126B (en) | Input method and electronic equipment | |
| CN109447258B (en) | Neural network model optimization method and device, electronic device and storage medium | |
| CN113139560A (en) | Training method and device of video processing model, and video processing method and device |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |