Advances, Systems and Applications
From: Data-intensive workflow scheduling strategy based on deep reinforcement learning in multi-clouds
Parameters | Value |
---|---|
Learning rate | 0.002 |
Reward attenuation rate | 0.9 |
Greed factor | 0.7 |
Max | 0.95 |
Growth rate | 1e-5 |
Experience pool storage size | 10,000 |
Minibatch | 128 |
replace_target_iter | 500 |
Activation function | Relu |
Hidden layer | 7 |