Gamma (γ)
Gamma (γ) is the parameter for discount factor.
Discount factor (gamma) represents how much the model cares about rewards in the future.
Example
When gamma is 0.9, the model will consider the reward in 6 steps half as important as immediate reward. When gamma is 0.95, the model will consider the reward in 13 steps half as important as immediate reward. When gamma is 0.99, the model will consider the reward in 60 steps half as important as immediate reward.
A model with gamma at 0.9 will seek more immediate rewards, where a model with gamma at 0.99 will seek more distant rewards.
Further readings
- Discount factor - Wikipedia
- What is the Full Meaning of the Discount Factor γ (gamma) in Reinforcement Learning? - Stack Overflow