Abstract
We propose R A PP, a new methodology for novelty detection by utilizing hidden space activation values obtained from a deep autoencoder. Precisely, R A PP compares input and its autoencoder reconstruction not only in the input space but also in the hidden spaces. We show that if we feed a reconstructed input to the same autoencoder again, its activated values in a hidden space are equivalent to the corresponding reconstruction in that hidden space given the original input. We devise two metrics aggregating those hidden activated values to quantify the novelty of the input. Through extensive experiments using diverse datasets, we validate that R A PP improves novelty detection performances of autoencoder-based approaches. Besides, we show that R A PP outperforms recent novelty detection methods evaluated on popular benchmarks.