Unsupervised Feature Learning via Non-Parametric Instance-level Discrimination

Abstract

Neural net classifiers trained on data with annotated class labels can also capture apparent visual similarity among categories without being directed to do so. We study whether this observation can be extended beyond the conventional domain of supervised learning: Can we learn a good feature representation that captures apparent similarity among instances, instead of classes, by merely asking the feature to be discriminative of individual instances? We formulate this intuition as a non-parametric classification problem at the instance-level, and use noise-contrastive estimation to tackle the computational challenges imposed by the large number of instance classes. Our experimental results demonstrate that, under unsupervised learning settings, our method surpasses the state-of-the-art on ImageNet classification by a large margin. Our method is also remarkable for consistently improving test performance with more training data and better network architectures. By fine-tuning the learned feature, we further obtain competitive results for semi-supervised learning and object detection tasks. Our non-parametric model is highly compact: With 128 features per image, our method requires only 600MB storage for a million images, enabling fast nearest neighbour retrieval at the run time.

CVPR 2018 spotlight paper. Code: https://github.com/zhirongw/lemniscate.pytorch

References in corpus (6)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Distilling the Knowledge in a Neural Network
• Prototypical Networks for Few-shot Learning
• SphereFace: Deep Hypersphere Embedding for Face Recognition
• Unsupervised Learning of Depth and Ego-Motion from Video
• Transitive Invariance for Self-supervised Visual Representation Learning

Cited by in corpus (41)

• Learning Transferable Visual Models From Natural Language Supervision
• Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval
• Contrastive Learning with Adversarial Examples
• Whitening for Self-Supervised Representation Learning
• Contrastive Model Inversion for Data-Free Knowledge Distillation
• Self-supervised Learning for Large-scale Item Recommendations
• MedAug: Contrastive learning leveraging patient metadata improves representations for chest X-ray interpretation
• Domain Adaptation for Semantic Segmentation via Patch-Wise Contrastive Learning
• Capturing the physics of MaNGA galaxies with self-supervised Machine Learning
• Viewmaker Networks: Learning Views for Unsupervised Representation Learning
• Hybrid Discriminative-Generative Training via Contrastive Learning
• A Large-Scale Study on Unsupervised Spatiotemporal Representation Learning
• Diversity Helps: Unsupervised Few-shot Learning via Distribution Shift-based Data Augmentation
• Adaptive Deep Kernel Learning
• Learning Visual Robotic Control Efficiently with Contrastive Pre-training and Data Augmentation
• SSFL: Tackling Label Deficiency in Federated Learning via Personalized Self-Supervision
• HarperValleyBank: A Domain-Specific Spoken Dialog Corpus
• Joint Contrastive Learning with Infinite Possibilities
• How Useful is Self-Supervised Pretraining for Visual Tasks?
• SegSort: Segmentation by Discriminative Sorting of Segments
• Multi-label Contrastive Predictive Coding
• Geometric Entropic Exploration
• Language-Driven Image Style Transfer
• OpenCoS: Contrastive Semi-supervised Learning for Handling Open-set Unlabeled Data
• For self-supervised learning, Rationality implies generalization, provably
• Calibrating and Improving Graph Contrastive Learning
• Contrastive Distillation on Intermediate Representations for Language Model Compression
• SimTriplet: Simple Triplet Representation Learning with a Single GPU
• KinePose: A temporally optimized inverse kinematics technique for 6DOF human pose estimation with biomechanical constraints
• Few-Shot Open-Set Recognition using Meta-Learning
• Self-supervised Visual Attribute Learning for Fashion Compatibility
• Label Contrastive Coding based Graph Neural Network for Graph Classification
• Self-Supervised Learning of a Biologically-Inspired Visual Texture Model
• DivCo: Diverse Conditional Image Synthesis via Contrastive Generative Adversarial Network
• Cluster Analysis with Deep Embeddings and Contrastive Learning
• Self-supervised Adversarial Training
• Multilayer Clustered Graph Learning
• Improving Context-Based Meta-Reinforcement Learning with Self-Supervised Trajectory Contrastive Learning
• Improving Object Detection with Selective Self-supervised Self-training
• CLAWS: Contrastive Learning with hard Attention and Weak Supervision
• Label-free Monitoring of Self-Supervised Learning Progress