Plan

1.DQN Adventure

• DQN
• Double DQN
• Dueling DQN
• DQN with prioritized experience replay
• Noisy nets
• Categorical DQN
• Rainbow DQN
• Quantile regression DQN
• Hierarchical DQN
• Neural Episodic Controle

2.RL Adventure

• Actor Critic
• ACKTR
• ACER
• PPO
• TRPO
• DDPG
• NAF
• Curiosity Driven Exploration
• TreeQN and ATreeC

RL-Adventure

1. a2c
2. gae
3. ppo
4. acer
5. sac
6. ddpg
7. dueling ddpg
8. her
9. gail

3.Imagination augmented agents

4.Play Snake with imagination

5.Transfer Tasks OpenAI

6.MetaTorch Library

7.NLP Adventure

8.Deep Graph Library

9.Deep Memory Library

План на МАРТ:

1. RL-Adventure-2
2. World Models
3. MetaRL

План

1. Опубликовать I2A
2. Опубликовать DQN Tutorial
3. Решать MetaRL
4. Решать рекурентные слои - conv,transformer,byte
5. Опубликовать Memory Networks
6. Заняться OpenAI Request

---------Долги

• RL adventure
• NLP adventure
• MetaTorch
• DeepGraph
• DeepMemory
• Anna Haritunyan papers Retrace -> Options, soft Q Learning, Udacity RL
• HRL papers
• Successor RL

План:

1. DeepMemory
2. Memory RL
3. Exploration
4. TDM
5. Levine prob RL
6. NLP adventure
7. MetaRL
8. DeepGraph