https://github.com/gunh0/reinforcement-learning-cartpole-balancing

📢 2019 Microsoft Student Partners (MSP) Evangelism Seminar - 2019.03.31
https://github.com/gunh0/reinforcement-learning-cartpole-balancing

artificial-intelligence cartpole microsoft-student-partners msp reinforcement-learning

Last synced: 8 months ago
JSON representation

📢 2019 Microsoft Student Partners (MSP) Evangelism Seminar - 2019.03.31

• Host: GitHub
• URL: https://github.com/gunh0/reinforcement-learning-cartpole-balancing
• Owner: gunh0
• License: mit
• Created: 2019-03-31T12:16:30.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2024-09-03T05:18:21.000Z (about 1 year ago)
• Last Synced: 2024-12-03T12:11:10.611Z (10 months ago)
• Topics: artificial-intelligence, cartpole, microsoft-student-partners, msp, reinforcement-learning
• Language: Jupyter Notebook
• Homepage:
• Size: 5.73 MB
• Stars: 2
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
### 2019 Microsoft Student Partners (MSP) Evangelism Seminar

**처음 시작하는 강화학습 with OpenAI Gym**

**2019. 03. 31**

![msp-logo.png](./docs/image/msp-logo.png)

---

**Cart Pole 균형 문제는 유전자 알고리즘, 인공신경망, 강화학습 등을 이용한 제어 전략 분야의 표준 문제이다.**

![cartpole-task.gif](/docs/image/cartpole-task.gif)

### Result (legacy)

![result-old.png](./docs/image/result-old.png)




### Last Updated (2024. 01.)

> 

- python 3.11.9

This tutorial shows how to use PyTorch to train a Deep Q Learning (DQN) agent on the CartPole-v1 task from Gymnasium.

![output.png](./docs/image/output.png)

**Diagram**

![diagram.png](./docs/image/diagram.jpg)

Actions are chosen either randomly or based on a policy, getting the next step sample from the gym environment. We record the results in the replay memory and also run optimization step on every iteration. Optimization picks a random batch from the replay memory to do training of the new policy. The “older” target_net is also used in optimization to compute the expected Q values. A soft update of its weights are performed at every step.