Semi-Supervised Learning with IPM-based GANs: an Empirical Study
This work addresses the challenge of enhancing semi-supervised learning efficiency for machine learning practitioners, but it is incremental as it builds on existing IPM-based GAN methods.
The paper tackled the problem of improving semi-supervised learning performance by empirically investigating how the critic design in IPM-based GANs influences results, finding that specific design choices like the K+1 formulation and avoiding batch normalization lead to better performance.
We present an empirical investigation of a recent class of Generative Adversarial Networks (GANs) using Integral Probability Metrics (IPM) and their performance for semi-supervised learning. IPM-based GANs like Wasserstein GAN, Fisher GAN and Sobolev GAN have desirable properties in terms of theoretical understanding, training stability, and a meaningful loss. In this work we investigate how the design of the critic (or discriminator) influences the performance in semi-supervised learning. We distill three key take-aways which are important for good SSL performance: (1) the K+1 formulation, (2) avoiding batch normalization in the critic and (3) avoiding gradient penalty constraints on the classification layer.