git clone https://github.com/davidwmcdevitt/MLDSRL
HW4 Submission - Genetic Algorithm and Bayesian Optimization
Genetic Algorithm
python MLDSRL/HW5/automl.py
Bayesian Optimization
python MLDSRL/HW5/bayesianoptimization.py
HW3.5 Submission - Differential Privacy
python MLDSRL/HW4/fedavg_differentialprivacy.py
Report: https://docs.google.com/document/d/120MVXAx2WUGw36rsecIOm8ktg8ULs5H9o3hg87ZZ2Lg/edit?usp=sharing
HW3 Submission - Federated Learning
Part 1
python MLDSRL/HW4/fedavg.py
Part 2
python MLDSRL/HW4/fedavg_distributed.py
Report: https://docs.google.com/document/d/1cgRGlaOCRhLXVbQCCI1M_jrQSrXmgreMIUHAYgwmn68/edit?usp=sharing
HW3 Submission
- CartPole Submission
python MLDSRL/HW3/cartpole_dqn.py
500 Episodes Rollout Test: Mean: 199.37 SD: 5.82
- Ms. Pacman Submission
python MLDSRL/HW2/mspacman_dqn.py
500 Episodes Rollout Test: Mean: 1803.74 SD: 669.7
HW2 Submission
- CartPole Submission
python MLDSRL/HW2/cartpole_base.py
CartPole trains until a streak of 10 consecutive episodes with a duration greater than or equal to 195 steps is acheived. Algorithm is tested on 100 episodes. if The average value of those 100 episodes is greater than or equal to 195, then the problem has been solved.
- Pong Submission
python MLDSRL/HW2/pong_baseline_v2.py
HW1 Submission
- CartPole Submission
python MLDSRL/HW1/cartpole.py
CartPole trains until a streak of 10 consecutive episodes with a duration greater than or equal to 195 steps is acheived. Algorithm is tested on 100 episodes. if The average value of those 100 episodes is greater than or equal to 195, then the problem has been solved.
- Pong Submission
python MLDSRL/HW1/pong.py
