Epsilon (ε) is probability of random movement.
It is used in the epsilon-greedy strategy to perform epsilon decay.
0 (0%) means no random moves, completely deterministic. 1 (100%) means completely random moves.
A model with high epsilon will make more random moves, this is useful for learning at the start. A model with low epsilon will make fewer random moves, this is useful for getting good result at the end.
- Epsilon Greedy in Deep Q Learning
- Epsilon-Greedy Q-learning
- Epsilon and learning rate decay in epsilon greedy q learning - Stack Overflow