Autoencoder Watchdog Outlier Detection for Classifiers
This addresses the issue of unreliable AI predictions for users in safety-critical applications, but it is incremental as it builds on existing autoencoder methods.
The paper tackles the problem of neural networks misclassifying out-of-distribution inputs by proposing an autoencoder watchdog to screen inputs before classification, with preliminary results demonstrated on MNIST images.
Neural networks have often been described as black boxes. A generic neural network trained to differentiate between kittens and puppies will classify a picture of a kumquat as a kitten or a puppy. An autoencoder watch dog screens trained classifier/regression machine input candidates before processing, e.g. to first test whether the neural network input is a puppy or a kitten. Preliminary results are presented using convolutional neural networks and convolutional autoencoder watchdogs using MNIST images.