Modeling Uncertainty in Learning with Little Data
Prof. Richard Zemel (Univ. of Toronto)
Few-shot classification, the task of adapting a classifier to unseen classes given a small labeled dataset, is an important step on the path toward human-like machine learning. I will present what I think are some of the key advances and open questions in this area. I will then focus on the fundamental issue of overfitting in the few-shot scenario. Bayesian methods are well-suited to tackling this issue because they allow practitioners to specify prior beliefs and update those beliefs in light of observed data. Contemporary approaches to Bayesian few-shot classification maintain a posterior distribution over model parameters, which is slow and requires storage that scales with model size. Instead, we propose a Gaussian process classifier based on a novel combination of Pólya-gamma augmentation and the one-vs-each loss that allows us to efficiently marginalize over functions rather than model parameters. We demonstrate improved accuracy and uncertainty quantification on both standard few-shot classification benchmarks and few-shot domain transfer tasks.
- - -
ML Seminars is a series of lectures on a variety of Machine Learning topics. Invited speakers for the series are leaders in their fields, hailing from respected research institutions worldwide. The ML Seminar Series is by the UT Austin Foundations of Data Science, an NSF Tripods Institute, and is administrated by WNCG.
[ Ссылка ]
