Adversarial Fisher Vectors for Unsupervised Representation Learning
This work addresses the challenge of extracting meaningful features from GANs for unsupervised learning, offering a novel approach that could benefit researchers in generative models and representation learning, though it appears incremental as it builds on existing GAN and EBM frameworks.
The paper tackles the problem of using GANs for unsupervised representation learning by reformulating them as energy-based models, showing that the discriminator can provide useful features for downstream tasks like classification and perceptual similarity, with competitive performance demonstrated in experiments.
We examine Generative Adversarial Networks (GANs) through the lens of deep Energy Based Models (EBMs), with the goal of exploiting the density model that follows from this formulation. In contrast to a traditional view where the discriminator learns a constant function when reaching convergence, here we show that it can provide useful information for downstream tasks, e.g., feature extraction for classification. To be concrete, in the EBM formulation, the discriminator learns an unnormalized density function (i.e., the negative energy term) that characterizes the data manifold. We propose to evaluate both the generator and the discriminator by deriving corresponding Fisher Score and Fisher Information from the EBM. We show that by assuming that the generated examples form an estimate of the learned density, both the Fisher Information and the normalized Fisher Vectors are easy to compute. We also show that we are able to derive a distance metric between examples and between sets of examples. We conduct experiments showing that the GAN-induced Fisher Vectors demonstrate competitive performance as unsupervised feature extractors for classification and perceptual similarity tasks. Code is available at \url{https://github.com/apple/ml-afv}.