---
title: "Discussion on \"Trust Region Methods: From REINFORCE to TRPO to PPO\" | Hashnode"
description: "Discussion on \"Trust Region Methods: From REINFORCE to TRPO to PPO\". Implement REINFORCE, TRPO, and PPO from scratch in PyTorch on CartPole. Understand why vanilla policy gradients are unstable, how trust regions fix it, and why PPO became the default algorithm in modern RL."
url: "https://hashnode.com/posts/trust-region-methods-from-reinforce-to-trpo-to-ppo/6a325fcda74418b00044d612"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Trust Region Methods: From REINFORCE to TRPO to PPO](https://sesenai.hashnode.dev/trust-region-methods-from-reinforce-to-trpo-to-ppo?utm_source=hashnode&utm_medium=feed)

Implement REINFORCE, TRPO, and PPO from scratch in PyTorch on CartPole. Understand why vanilla policy gradients are unstable, how trust regions fix it, and why PPO became the default algorithm in modern RL.

Jun 17 · 21 min read · sesenai.hashnode.dev  
Read the article

-   [#reinforcement-learning](https://hashnode.com/tag/reinforcement-learning)
-   [#deep-learning](https://hashnode.com/tag/deep-learning)
-   [#optimisation](https://hashnode.com/tag/optimisation)

## Written by

[Berkan Sesen](https://hashnode.com/@berkansesen)

@berkansesen · London, UK

  
[View profile](https://hashnode.com/@berkansesen)

I think about AI. A lot.

I think about AI. A lot.

## Responses

Markdown supported · ⌘↵ to post

No responses yet.