Curriculum learning is a technique inmachine learning in which amodel is trained on examples of increasing difficulty, where the definition of "difficulty" may be provided externally or discovered as part of the training process. This is intended to attain good performance more quickly, or to converge to a better local optimum if theglobal optimum is not found.^[1][2]

Most generally, curriculum learning is the technique of successively increasing the difficulty of examples in thetraining set that is presented to a model over multiple training iterations. This can produce better results than exposing the model to the full training set immediately under some circumstances; most typically, when the model is able to learn general principles from easier examples, and then gradually incorporate more complex and nuanced information as harder examples are introduced, such asedge cases. This has been shown to work in many domains, most likely as a form ofregularization.^[3]

There are several major variations in how the technique is applied:

• A concept of "difficulty" must be defined. This may come from human annotation^[4][5] or an externalheuristic; for example inlanguage modeling, shorter sentences might be classified as easier than longer ones.^[6] Another approach is to use the performance of another model, with examples accurately predicted by that model being classified as easier (providing a connection toboosting).
• Difficulty can be increased steadily^[7] or in distinct epochs,^[8] and in a deterministic schedule or according to aprobability distribution. This may also be moderated by a requirement for diversity at each stage, in cases where easier examples are likely to be disproportionately similar to each other.^[9]
• Applications must also decide the schedule for increasing the difficulty. Simple approaches may use a fixed schedule, such as training on easy examples for half of the available iterations and then all examples for the second half.^[3] Other approaches useself-paced learning to increase the difficulty in proportion to the performance of the model on the current set.^[10]

Since curriculum learning only concerns the selection and ordering of training data, it can be combined with many other techniques in machine learning. The success of the method assumes that a model trained for an easier version of the problem cangeneralize to harder versions, so it can be seen as a form oftransfer learning. Some authors also consider curriculum learning to include other forms of progressively increasing complexity, such as increasing the number of model parameters.^[11] It is frequently combined withreinforcement learning, such as learning a simplified version of a game first.^[12]

Some domains have shown success withanti-curriculum learning: training on the most difficult examples first. One example is the ACCAN method forspeech recognition, which trains on the examples with the lowestsignal-to-noise ratio first.^[13]

The term "curriculum learning" was introduced byYoshua Bengio et al in 2009,^[14] with reference to thepsychological technique ofshaping in animals and structured education for humans: beginning with the simplest concepts and then building on them. The authors also note that the application of this technique in machine learning has its roots in the early study ofneural networks such asJeffrey Elman's 1993 paperLearning and development in neural networks: the importance of starting small.^[15] Bengio et al showed good results for problems inimage classification, such as identifyinggeometric shapes with progressively more complex forms, andlanguage modeling, such as training with a gradually expandingvocabulary. They conclude that, for curriculum strategies, "their beneficial effect is most pronounced on the testset", suggesting good generalization.

The technique has since been applied to many other domains:

• Natural language processing:
  • Part-of-speech tagging^[16]
  • Intent detection^[17]
  • Sentiment analysis^[18]
  • Machine translation^[19][20]
  • Speech recognition^[21]
  • Language model pre-training^[22]
• Image recognition:
  • Facial recognition^[23]
  • Object detection^[24]
• Reinforcement learning:
  • Game-playing^[25]
• Graph learning^[26][27]
• Matrix factorization^[28]

• Curriculum Learning: A Survey
• A Survey on Curriculum Learning
• Curriculum Learning for Reinforcement Learning Domains: A Framework and Survey
• Curriculum learning at IEEE Xplore

• Machine learning algorithms

• Articles with short description
• Short description matches Wikidata