Solutions to the multi-armed bandits problems described in book Reinforcement Learning: An Introduction by Sutton and Barto.
Includes the following algorithms for learning optimal strategies (see Agents.py)-
- epsilon greedy algorithm
- softmax algorithm
![/graphs/nArmedBanditAvgRewardsComparison.png](/graphs/nArmedBanditAvgRewardsComparison.png?raw=true "varying epsilon in epsilon greedy: avg. reward vs iterations") ![/graphs/eGreedyvsSoftmax.png](/graphs/eGreedyvsSoftmax.png?raw=true "Epsilon-Greedy vs Softmax performance")