Using Representation Expressiveness and Learnability to Evaluate Self-Supervised Learning Methods
This provides a label-free evaluation framework for SSL models, which is incremental as it builds on existing metrics to enhance correlation and transfer prediction.
The paper tackles the problem of evaluating self-supervised learning models without supervised labels by proposing CLID, a method combining expressiveness (via Intrinsic Dimension) and learnability (via Cluster Learnability). They find that CLID better correlates with in-distribution performance and improves out-of-domain generalization predictions compared to baselines.
We address the problem of evaluating the quality of self-supervised learning (SSL) models without access to supervised labels, while being agnostic to the architecture, learning algorithm or data manipulation used during training. We argue that representations can be evaluated through the lens of expressiveness and learnability. We propose to use the Intrinsic Dimension (ID) to assess expressiveness and introduce Cluster Learnability (CL) to assess learnability. CL is measured in terms of the performance of a KNN classifier trained to predict labels obtained by clustering the representations with K-means. We thus combine CL and ID into a single predictor -- CLID. Through a large-scale empirical study with a diverse family of SSL algorithms, we find that CLID better correlates with in-distribution model performance than other competing recent evaluation schemes. We also benchmark CLID on out-of-domain generalization, where CLID serves as a predictor of the transfer performance of SSL models on several visual classification tasks, yielding improvements with respect to the competing baselines.