|
Learning-based AI approach | Data-based learning approach |
Unsupervised learning (UL) for unlabeled data | Supervised learning (SL) for labeled data | Reinforcement learning (RL) for rewarded labeled data (labeled data with cost function) |
|
Cortical learning (CL): neuroscience of brain cortex (cortical areas) | Sequence learning: hierarchical temporal memory (HTM) and cortical learning algorithm (CLA) | — | — |
Machine learning (ML): shallow ML | Dimension reduction: principle component analysis (PCA) and independent component analysis (ICA), clustering: expectation maximization (EM), K-means, K-nearest neighbors (KNN), approximate nearest neighbor (ANN), and fast library for approximate nearest neighbor (FLANN) | Linear discriminant analysis (LDA), random forest, search trees (Monte Carlo search), artificial neural network (ANN) or multilayer perceptron (MLP), and support vector machine (SVM) | Q-learning, policy-learning, and inverse RL (IRL) |
Deep learning (DL): deep ML | Deep unsupervised learning (DUL): restricted Boltzmann machine (RBM), deep belief network (DBN), deep Boltzmann machine (DBM), autoencoder (AE), variational autoencoder (VAE), generative adversarial network (GAN), and sequence learning | Deep supervised learning (DSL): Feed-forward neural network (FFNN), deep neural network (DNN), spike neural network (SNN), sequence-to-sequence learning, recurrent neural network (vanilla RNN), long short-term memory (LSTM), convolutional LSTM (ConvLSTM), and gated recurrent unit (GRU) | Deep reinforcement learning (DRL): Deep Q-Network (DQN), AlphaGo, and inverse DRL (inverse RL & GAN) |
|