Tri-contrastive Learning: Identifiable Representation Learning with Automatic Feature Importance Discovery
Dec 10, 2023
Speakers
Yifei Wang
Speaker · 0 followers
Yifei Wang
Speaker · 0 followers
About
Existing contrastive learning methods rely on pairwise sample contrast z_x^⊤ z_x' to learn data representations, but the learned features often lack clear interpretability from a human perspective. Theoretically, it lacks feature identifiability and different initialization may lead to totally different features.In this paper, we study a new method named tri-contrastive learning (triCL) that involves a 3-factor contrast in the form of z_x^⊤ S z_x', where S= diag(s_1,...,s_k) is a learnable diagonal matrix that automatically captures the importance of each feature. We show that by this simple extension, triCL can not only obtain identifiable features that eliminate randomness, but also obtain more interpretable features that are ordered according to the importance matrix S. We show that features with high importance have nice interpretability by capturing common classwise features, and obtain superior performance when evaluated for image retrieval using a few features. The proposed triCL objective is general and can be applied to different contrastive learning methods like SimCLR and CLIP. We believe that it is a better alternative to existing 2-factor contrastive learning by improving its identifiability and interpretability with minimal overhead.Existing contrastive learning methods rely on pairwise sample contrast z_x^⊤ z_x' to learn data representations, but the learned features often lack clear interpretability from a human perspective. Theoretically, it lacks feature identifiability and different initialization may lead to totally different features.In this paper, we study a new method named tri-contrastive learning (triCL) that involves a 3-factor contrast in the form of z_x^⊤ S z_x', where S= diag(s_1,...,s_k) is a learnable diago…
Organizer
NeurIPS 2023
Account · 617 followers
Like the format? Trust SlidesLive to capture your next event!
Professional recording and live streaming, delivered globally.
Sharing
Recommended Videos
Presentations on similar topic, category or speaker
Conformal Prediction for Uncertainty-Aware Planning with Diffusion Dynamics Model
Watch later
Jiankai Sun, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Federated Linear Bandits with Finite Adversarial Actions
Watch later
Li Fan, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
On Evaluating Adversarial Robustness of Large Vision-Language Models
Watch later
Yunqing Zhao, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Skilful Precipitation Nowcasting Using NowcastNet
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Grassmann Manifold Flows for Stable Shape Generation
Watch later
Ryoma Yataka, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Transfer Learning, Reinforcement Learning for Adaptive Control Optimization under Distribution Shift
Watch later
Pankaj Rajak, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%