Skip to main content

Table 2 Hyper parameters of the LSTM encoder-decoder network with attention

From: A novel approach to workload prediction using attention-based LSTM encoder-decoder network in cloud environment

Hyper parameter

Value

History window length

18

Dimension of hidden state in encoder

64

Dimension of hidden state in decoder

64

Batch size

128

Factor for momentum μ

0.8

δ in Huber loss function

1.35

Initial learning rate η

0.001

Factor for moving average β1

0.9

Factor for moving average β2

0.999