REVIEW 4 cited by
What Should Not Be Contrastive in Contrastive Learning
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Recent self-supervised contrastive methods have been able to produce impressive transferable visual representations by learning to be invariant to different data augmentations. However, these methods implicitly assume a particular set of representational invariances (e.g., invariance to color), and can perform poorly when a downstream task violates this assumption (e.g., distinguishing red vs. yellow cars). We introduce a contrastive learning framework which does not require prior knowledge of specific, task-dependent invariances. Our model learns to capture varying and invariant factors for visual representations by constructing separate embedding spaces, each of which is invariant to all but one augmentation. We use a multi-head network with a shared backbone which captures information across each augmentation and alone outperforms all baselines on downstream tasks. We further find that the concatenation of the invariant and varying spaces performs best across all tasks we investigate, including coarse-grained, fine-grained, and few-shot downstream classification tasks, and various data corruptions.
Forward citations
Cited by 4 Pith papers
-
Self-Supervised Dynamical System Representations for Physiological Time-Series
PULSE pretrains physiological time-series encoders by reconstructing random crops from inferred system parameters, improving label efficiency and transfer across four sensor domains.
-
FRAME: Pre-Training Video Feature Representations via Anticipation and Memory
FRAME distills DINO and CLIP features into a compact video encoder with a memory module and future-frame prediction, outperforming image-based and self-supervised video baselines on dense video tasks.
-
The Impact of Semantic Pairs on Self-Supervised Representation Learning
Semantic positive pairs (same-class, different images) consistently beat augmented-view pairs for self-supervised pretraining in controlled ImageNet-subset experiments, with the largest gains for SimCLR.
-
Directed Link Prediction using GNN with Local and Global Feature Fusion
A GNN that fuses path labels, community labels, and contrastive embeddings on a directed line graph is reported to beat six baselines on directed link prediction, but its proof has errors.
Discussion (0). Sign in to comment.