Artificial Intelligence (AI) Basics

Indramal Wansekara

5 min read

·

Jul 24, 2021

--

Listen

Share

Press enter or click to view image in full size

Summary: This article wrote to get a basic understanding of artificial intelligence (AI). When you get an overall basic understanding knowledge of AI, you can consider what should you learn, and which AI area should you use for your applications because AI has many areas and it is used in many applications to solve some kinds of problems (fully or partially).

Artificial intelligence (AI) is atoolthat can be used in various types of applications. Mostly AI application writes using Python language or R language. If you are planning to learn or use AI, you should learn Python or R language. But sometimes some AI specialists use other languages like .Net, Javascript (JS), and so on. When we use other languages, AI tools can use in many applications.

• ML.Net —Link — This is .Net framework to use with .Net applications.
• TensorFlow.js —Link — This is a javascript version of Google Tensorflow. It can use on website applications.

👉 You should understand this, AI uses for predictions and not to get 100% accurate results. When we code a program for some AI application, we call it a Model. Depending on the model's accuracy, the result will vary. When a model predition accuracy is high, you will get best results to your application.

Let's break down this article into the following topics.

• Artificial Intelligent (AI) methods breakdown
• Most use Artificial Intelligent (AI) applications and use AI methods
• Available python frameworks and mostly use python packages
• Find Dataset for your AI model
• Freely available online platforms for AI coding
• Available hardware for AI applications

Before discussing the above topics, I think you need to know what the Dataset is. Dataset is a collection of data. It can be text data, image data, video data, or audio data. Dataset uses to train and test a model. Dataset data should be well organized to insert into the model. Some datasets have labels. This means those labels say what data are these. Example: If you have apple fruit images data, the label has it is apple. Mostly the labels can be folder names or a CVS file with the list of labels. Also can be other arrangements like a text file etc. It is dependent on the dataset creator.

📚Artificial Intelligent (AI) methods breakdown

We can divide AI into Classical Methods and Modern Methods.

📕 Classical Methods:

Classical methods are used in begging of AI development. Those methods are still used but have less accuracy and can use for one thing. You can google the following methods name to get more information.

These are methods,

1. Regression(Examples: Simple Linear Regression, Polynomial Regression, Support Vector Regression (SVR), Decision Tree Regression)
2. Classification(Examples: Logistic Regression, K-Nearest Neighbor (K-NN), Support Vector Machine (SVM), Naive Bayes, Random Forest)
3. Clustering(Examples: K-Means Clustering, Hierarchical Clustering)

📗 Modern Methods:

Modern methods are widely used for AI applications. Modern methods can be divided into three different sections.

• Unsupervised Learning — This method uses not labeled datasets.
• Supervised Learning — This method uses labeled datasets.
• Reinforcement Learning — This method does not use any datasets. This type of model learns itself by experiencing.

Unsupervised Learning

These are unsupervised learning methods,

1. Self Organizing Maps
2. Boltzmann Machine
3. AutoEnccoders

Supervised Learning

These are supervised learning methods,

1. Deep Neural Networks (DNN)
2. Convolutional Neural Networks (CNN)
3. Recurrent Neural Networks (RNN)
4. Generative Adversarial Network (GAN)
5. Transformer Neural Networks

Reinforcement Learning

These are reinforcement learning methods,

1. Q Learning
2. DQN (Q-Learning + deep neural networks)
3. Deep Deterministic Policy Gradient (DDPG)
4. Asynchronous Advantage Actor Critic (A3C)
5. Advantage Actor Critic (A2C)

There are a lot of Reinforcement learning methods available now and divided into several sections according to learning type of reinforcement learning. The above methods should be under one of those sections.

The below photo shows these sections' taxonomy for some reinforcement learning methods.

There are a lot of models created by using the above methods. You can just use a trained model for your application. You can check which dataset is used when the model is training. Because you need to know whether your chosen model can or can not predict results according to your application. Example: If you choose a Convolutional Neural Network (CNN) method that used an image recognition model to predict or identify an apple image but while training the model by some other model creator not used the apple images as data, you can not use this model to predict or identify an apple.

📚 Most use Artificial Intelligent (AI) applications and use AI methods

:orange_book: Computer Vision Application

AI can use for this applications and there are main areas in computer vision applications.

• Classification
• Classification + Localisation
• Object Detection
• Segmentation

Press enter or click to view image in full size

Use AI methods to create a models

• Convolutional Neural Network (CNN) + Deep Neural Network (DNN)
• Generative Adversarial Network (GAN)

📘 Natural Language Processing (NLP)

AI can use for this application to understand language. There are main areas in Natural Language Processing.

• Translations
• Speech-to-text conversations
• Text generation
• Questions and Answering
• Image capturing

Use AI methods to create a models

• Recurrent Neural Networks (RNN) + Deep Neural Networks (DNN)
• Transformer Neural Networks

I will update this article from time to time.

📚 Available python frameworks and mostly use python packages

Coming soon.

📚 Find Dataset for your AI model

Coming soon

📚 Freely available online platforms for AI coding

Coming soon

📚 Available hardware for AI applications

When we consider hardware for AI programs, there are few companies making suitable hardware like Nvidia, Google, and AMD.

— — — — — — — — — — — — — — — — —

Please do not worry about the above-mentioned words or names. You can google those words and check for more information. I hope you get a basic idea of AI. If you have any comments, please enter it in the comment section. Your support is highly appreciated.