📰 Paper reading club: PPO experience report
postponed for a week due to the wind!

We come together to discuss advances in computer science, ground them down to applications, and see where to the wind is blowing. Read the papers at your own time and come discuss them with us.

Primacy bias in PPO
"The Primacy Bias in RL" shows how we can improve sample efficiency and increase the expected return with weight resets. I will tell how we tried leveraging resets in PPO and analyzed primacy bias phenomena.

The plan:
- How to compare RL algorithms: "rliable" recommendations
- The Primacy Bias in RL: when resets work, and when they don't
- Resets in PPO: what we tried (spoiler: nothing worked)
- Heavy Priming experiment: overfitting on one batch at the beginning of the training. Differences between SAC and PPO, the probable cause of PPO stability.

⏱ 18:00-20:00 Thursday, 28 December
📍 F0RTHSP4CE, Khorava St, 18
💰 free/donation
🗣 @iwksx (🇬🇧/🇷🇺)
👮 @cad215, @to_be_elsewhere