202509281334
Status: #idea
Tags: #reinforcement_learning #ai
# Why we use a discount factor in reinforcement learning
There are two core reasons why a discount factor is needed in RL:
1. The later we receive rewards, the less attractive they are to present calculations
2. To reduce the variance of return estimates. Given the future is uncertain, the further we look into the future, the more we accumulate random variance and the more variance our value estimates have.
---
# References
[[Grokking Deep Reinforcement Learning]]