Transferability and Hardness of Supervised Classification Tasks
This work addresses the challenge of predicting task performance and transferability in machine learning, which is incremental as it builds on prior transfer learning methods by introducing a solution-agnostic approach.
The authors tackled the problem of estimating task difficulty and transferability for supervised classification without needing trained models, using an information-theoretic approach based on label statistics, and demonstrated strong correlations with empirical results across 437 tasks, achieving state-of-the-art accuracy in a face recognition transfer case study.
We propose a novel approach for estimating the difficulty and transferability of supervised classification tasks. Unlike previous work, our approach is solution agnostic and does not require or assume trained models. Instead, we estimate these values using an information theoretic approach: treating training labels as random variables and exploring their statistics. When transferring from a source to a target task, we consider the conditional entropy between two such variables (i.e., label assignments of the two tasks). We show analytically and empirically that this value is related to the loss of the transferred model. We further show how to use this value to estimate task hardness. We test our claims extensively on three large scale data sets -- CelebA (40 tasks), Animals with Attributes 2 (85 tasks), and Caltech-UCSD Birds 200 (312 tasks) -- together representing 437 classification tasks. We provide results showing that our hardness and transferability estimates are strongly correlated with empirical hardness and transferability. As a case study, we transfer a learned face recognition model to CelebA attribute classification tasks, showing state of the art accuracy for tasks estimated to be highly transferable.