-
Reinforcement Learning #1
1. Mathematical Background - Probability - ok - Random Variable - ok - Random Process - ok 2. Basic RL Algorithm - Markov Process - ok - Markov Reward Process - ok - Markov Decision Process - ok - Bellman Expectation Eqn - Bellman Optimality Eqn - Dynamic Programming - ok - Value iteration - ok - Policy iteration - ok - Model free Approaches # MF Prediction - Monte Carlo - ok - Temporal Difference - ok (Example : Random Walk) # MF Control - Sarsa - ok - Q-Learning - ok (Example : Cliff Walking) (Example : Windy Grid) (Example : Windy Cliff) 3. ML based R learning - Function Approximation - ok - DQN - ok (Example : Cartpole - DQN) 4. Policy Base R Learning - REINFORCE - ok - A2C - ok # Term Project Cartpole A2C Cartpole DQN Cartpole REINFORCE
-
Reinforcement Learning #2
-
Week 1 : Dynamic Programming
- Policy Iteration - Value Iteration # Proof of Convergence
-
Week 2 : Monte Carlo
- On Policy Monte Carlo : Batch / Recursive - Off Policy Monte Carlo : Batch / Recursive # Law of Large Number # Empirical Mean # Importance Sampling
-
Week 3 : Temporal Difference
- Temporal difference(0) - Temporal difference(1) - Temporal difference(λ) - SARSA - Q Learning - Double Q Learning - Deep Q Learning - Function Approximation # Robbins-Monro rule # Sherman-Morrison fomular # Projected Bellman Eqn # Maximization bias
-
Week 4 : Policy Gradient
- REINFORCE - A2C - DPG - DDPG # PG Proof # Information Theory - Self Information - Shannon-Entropy - KL divergence - Cross Entropy
-
-
Week 5 : Advanced RL
- D3QN - Double Deep - Dueling Deep - TD3 - TRPO - PPO
-
Week 6 : Project
# Solve BiPedal
-
Machine Learning
1. Linear Regression - ok 2. Logistic Regresssion - ok 3. K-nearest neighborhood - ok 4. K-means clustering - ok 5. Naive Bayes - ok 6. SVM - ok 7. PCA 8. Decision Tree - ok 9. Perceptron - ok 1. SLP - ok 2. MLP - ok
-
Deep Learning
1. Linear Regression - ok 2. Logistic Regression - Logistic Regression(Binary Classification) - ok - Softmax Regression(MultiClass Classification) - ok 3. Auto Encoder - AE - ok - CAE - ok 4. Modern CNN - LeNet - ok - AlexNet - ok - VGG Nets - ok - GoogLeNet - ok - ResNet - ok 5. Semantic Segmentation - FCN - ok - DeConvNet - ok - SegNet - ok - U-Net - ok - DeepLab v1, v2 - ok 6. Object Detection - RCNN - Fast RCNN - Faster RCNN - SPP Net - Yolo - SDD - Attention Net 7. NLP - RNN - ok - LSTM / GRU - ok - Sequence Prediction - ok - Sequence Classification - ok