An LSTM-Autoencoder Architecture for Anomaly Detection Applied on Compressors Audio Data

Table 2

Model dimensions and operations overview for LSTM-AE architecture (created by authors).

Operation		Data	Dimensions	Weights ()	Weights (%)
Input	#####	2	259
LSTM	LLLLL	—	—	62,000	66.7%
tanh	#####		50
Dropout	|||	—	—	0	0.0%
	#####		50
Repeat vector		—	—	0	0.0%
	#####	259	50
LSTM	LLLLL	—	—	20,200	21.7%
tanh	#####	259	50
Dense	|||	—	—	0	0.0%
	#####	259	50
Time distributed		—	—	102	0.1. %
	#####	259	2
LSTM	LLLLL	—	—	10,600	11.4%
tanh	#####		50
Dense	XXXXX	—	—	102	0.1%
SoftMax	#####		2