AdversarialTensors implements a variety of tensor factorization methods for defending Artificial intelligence
(AI) models against adversarial attacks. The library implements three main operations. First, tensor
factorization methods are implemented as a preprocessing stage for input data to AI models to reduce the
effectiveness of adversarial noise. In the second operation, tensor factorization methods are used to
find novel latent attack features by combining proposed attacks from a variety of methods. Since these
attacks will inherently be a combination of attacks many algorithms against many models, they have the
potential to threaten a wide variety of AI models simultaneously In the third operation, an unsupervised
generative adversarial networks (GAN) is employed to generate denoised data from many adversarial noises.
This generator provides robust defense against unseen attacks.