VTGAN: hybrid generative adversarial networks for cloud workload prediction

Maiyza, Aya I.; Korany, Noha O.; Banawan, Karim; Hassan, Hanan A.; Sheta, Walaa M.

doi:10.1186/s13677-023-00473-z

Journal of Cloud Computing

Advances, Systems and Applications

Table 4 The structure of VTGAN models

From: VTGAN: hybrid generative adversarial networks for cloud workload prediction

Model	Layers	Configuration
Stacked LSTM	Bidirectional cuDNNLSTM	256 units, dropout= 0.2
		L1 kernel, recurrent, and bias regularization = 0.00001
	cuDNNLSTM	128 units, dropout= 0.2
		L1 kernel, recurrent, and bias regularization = 0.00001
	cuDNNLSTM	128 units, dropout= 0.2
		L1 kernel, recurrent, and bias regularization = 0.00001
	FC	Dense, output units (1 or p for one or multiple-step-ahead)
		L1 kernel, and bias regularization = 0.00001
Stacked GRU	Bidirectional cuDNNGRU	256 units, dropout= 0.2
		L1 kernel, recurrent, and bias regularization = 0.00001
	cuDNNGRU	128 units, dropout= 0.2
		L1 kernel, recurrent, and bias regularization = 0.00001
	cuDNNGRU	128 units, dropout= 0.2
		L1 kernel, recurrent, and bias regularization = 0.00001
	FC	Dense, output units (1 or p for one or multiple-step-ahead)
		L1 kernel, and bias regularization = 0.00001
VTGAN (LSTM-based)	- Generator parameters
	Bidirectional cuDNNLSTM	256 units, dropout= 0.2
		L1 kernel, recurrent, and bias regularization = 0.00001
	cuDNNLSTM	128 units, dropout= 0.2
		L1 kernel, recurrent, and bias regularization = 0.00001
	cuDNNLSTM	128 units, dropout= 0.2
		L1 kernel, recurrent, and bias regularization = 0.00001
	FC	Dense, output units (1 or p for one or multiple-step-ahead)
		L1 kernel, and bias regularization = 0.00001
	- Discriminator parameters
	Conv1D	flter=64, kernel size=5, strides=2, padding=same
		LeakyReLU activation (alpha=0.001)
	Conv1D	flter=128, kernel size=5, strides=2, padding=same
		LeakyReLU activation (alpha=0.001)
	Conv1D	flter=128, kernel size=5, strides=2, padding=same
		LeakyReLU activation (alpha=0.001)
	Flatten
	FC 1	Dense, units=64, LeakyReLU activation
	FC 2	Dense, output units (1 or p for one or multiple-step-ahead),
		sigmoid activation
VTGAN (GRU-based)	- Generator parameters
	Bidirectional cuDNNGRU	256 units, dropout= 0.2
		L1 kernel, recurrent, and bias regularization = 0.00001
	cuDNNGRU	128 units, dropout= 0.2
		L1 kernel, recurrent, and bias regularization = 0.00001
	cuDNNGRU	128 units, dropout= 0.2
		L1 kernel, recurrent, and bias regularization = 0.00001
	FC	Dense, output units (1 or p for one or multiple-step-ahead)
		L1 kernel, and bias regularization = 0.00001
	- Discriminator parameters:
	as VTGAN (LSTM-based)

Back to article page