policy-gradient-importance-sampling

KO | EN

gitlite — search

#javascript #python #react #ai #typescript #llm #rust #linux #docker #windows #awesome #awesome-list #ios #mcp

policy-gradient-importance-sampling

(★ 33)

Policy gradient reinforcement learning algorithm with importance sampling

policy-gradient-importance-sampling 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# policy-gradient-importance-sampling

Importance sampling is technique to estimate expectation of function under distribution p(x) with samples drawn from another distribution q(x). In policy gradient, this technique makes agent to use off-policy samples (samples T = (s1,a1,s2,a2,....) drawn from old policy) to update current policy. Thus, importance sampled policy gradient can reuses previous samples for training, ensures faster convergence.

Experiment on CartPole-v0 of [gym](https://github.com/openai/gym) shows importance sampling requires fewer episodes to learn optimal policy.

## Requirements

- Python
- [numpy](https://www.numpy.org/)
- [pytorch](https://pytorch.org/)
- [gym](https://github.com/openai/gym)
- [matplotlib](https://matplotlib.org/users/installing.html)

## Usage

$ python main.py (--reinforce to run REINFORCE example)

## Reference

[JamesChuangg's REINFORCE implementation](https://github.com/JamesChuanggg/pytorch-REINFORCE)

policy-gradient-importance-sampling

policy-gradient-importance-sampling 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

policy-gradient-importance-sampling

policy-gradient-importance-sampling 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README