The disclosed technology relates generally to outlier detection, and more particularly to deep learning neural networks and apparatuses and methods for detecting outliers using non-transitory computer-readable storage media. According to an embodiment, a method for detecting outliers using a deep learning neural network model includes providing a deep learning neural network model. A deep learning neural network model includes an encoder including a plurality of encoder layers and a decoder including a plurality of decoder layers. The method includes feeding a first input to the encoder to produce a first encoded input and continuously processing the first input through the plurality of encoder layers - continuously processing the first input further comprising generating a first intermediate encoded input from one of the encoder layers prior to generating the first encoded input. The method further comprises feeding the first encoded input from the encoder to the decoder and continuously processing the first encoded input through the plurality of decoder layers to produce a first reconstructed output. . The method comprises feeding the first reconstructed output from the decoder to the encoder as a second or next input and continuously processing the first reconstructed output through the plurality of encoder layers - the first reconstructed output Continuously processing the output further comprises - generating a second intermediate encoded input from one of the encoder layers. The method further comprises detecting an outlier in the original input based on a comparison of the first intermediate encoded input and the second intermediate encoded input.
展开▼
