Open-Set Recognition with Gaussian Mixture Variational Autoencoders

In inference, open-set classification is to either classify a sample into known class from training or reject it as an unknown class. Existing deep classifiers train explicit closed-set classifiers, in some cases disjointly utilizing reconstruction, which we find dilutes the latent representation's ability distinguish classes. contrast, our model cooperatively learn reconstruction and perform class-based clustering space. With this, Gaussian mixture variational autoencoder (GMVAE) achieves more accurate robust results, with average F1 increase of 0.26, through extensive experiments aided by analytical results.