Table 2 Best Hyper-parameters found during the validation procedure for the DQN algorithm.
From: A Deep Q-Network based hand gesture recognition system for control of robotic platforms
Hyper-parameter name | Hyper-parameter values |
|---|---|
Activation function between layers | Relu |
Target Smooth Factor | \(0.5e^{-3}\) |
Experience buffer length | 60 |
Learn rate (\(\alpha\)) | \(0.07e^{-3}\) |
Epsilon initial value | 0.99 |
Epsilon greedy epsilon decay | \(1.5e^{-4}\) |
Discount factor | 0.99 |
Mini batch size | 64 |
Optimizer | Adam |
Gradient decay factor | 0.85 |
L2 regularization factor | 0.0003 |
