Twin Contrastive Learning for Online Clustering (2022-07-09T00:00:00.000000Z)

TL;DR

It is found that when the data is projected into a feature space with a dimensionality of the target cluster number, the rows and columns of its feature matrix correspond to the instance and cluster representation, respectively.

Abstract

This paper proposes to perform online clustering by conducting twin contrastive learning (TCL) at the instance and cluster level. Specifically, we find that when the data is projected into a feature space with a dimensionality of the target cluster number, the rows and columns of its feature matrix correspond to the instance and cluster representation, respectively. Based on the observation, for a given dataset, the proposed TCL first constructs positive and negative pairs through data augmentations. Thereafter, in the row and column space of the feature matrix, instance- and cluster-level contrastive learning are respectively conducted by pulling together positive pairs while pushing apart the negatives. To alleviate the influence of intrinsic false-negative pairs and rectify cluster assignments, we adopt a confidence-based criterion to select pseudo-labels for boosting both the instance- and cluster-level contrastive learning. As a result, the clustering performance is further improved. Besides the elegant idea of twin contrastive learning, another advantage of TCL is that it could independently predict the cluster assignment for each instance, thus effortlessly fitting online scenarios. Extensive experiments on six widely-used image and text benchmarks demonstrate the effectiveness of TCL. The code is released on https://pengxi.me.

Authors

Mouxing Yang

4 papers

Yunfan Li

4 papers

Xiaocui Peng

4 papers

TL;DR

Abstract

Authors

References93 items

Revisiting Contrastive Methods for Unsupervised Learning of Visual Representations

With a Little Help from My Friends: Nearest-Neighbor Contrastive Learning of Visual Representations

Contrastive Learning With Stronger Augmentations

Supporting Clustering with Contrastive Learning

Structure-Aware Face Clustering on a Large-Scale Graph with 107 Nodes

SPICE: Semantic Pseudo-Labeling for Image Clustering

Doubly Contrastive Deep Clustering

Barlow Twins: Self-Supervised Learning via Redundancy Reduction

A statistical interpretation of term specificity and its application in retrieval

Improving Unsupervised Image Clustering With Robust Learning

Exploring Simple Siamese Representation Learning

AdCo: Adversarial Contrast for Efficient Learning of Unsupervised Representations from Self-Trained Negative Adversaries

Unsupervised Feature Learning by Cross-Level Instance-Group Discrimination

Deep Robust Clustering by Contrastive Learning

Unsupervised Learning of Visual Features by Contrasting Cluster Assignments

Big Self-Supervised Models are Strong Semi-Supervised Learners

Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning

Deep Semantic Clustering by Partition Confidence Maximisation

Learning To Classify Images Without Labels

Prototypical Contrastive Learning of Unsupervised Representations

Supervised Contrastive Learning

Autoencoder Constrained Clustering With Adaptive Neighbors

Improved Baselines with Momentum Contrastive Learning

A Simple Framework for Contrastive Learning of Visual Representations

Enhancement of Short Text Clustering by Iterative Classification

Deep Clustering With Sample-Assignment Invariance Prior

PyTorch: An Imperative Style, High-Performance Deep Learning Library

Momentum Contrast for Unsupervised Visual Representation Learning

Self-labelling via simultaneous clustering and representation learning

Randaugment: Practical automated data augmentation with a reduced search space

NLNL: Negative Learning for Noisy Labels

Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks

K-Multiple-Means: A Multiple-Means Clustering Method with Specified K Clusters

Deep Discriminative Clustering Analysis

Deep Comprehensive Correlation Mining for Image Clustering

EDA: Easy Data Augmentation Techniques for Boosting Performance on Text Classification Tasks

Challenges in unsupervised clustering of single-cell RNA-seq data

Associative Deep Clustering: Training a Classification Network with No Labels

Kernel Cuts: Kernel and Spectral Clustering Meet Regularization

XAI Beyond Classification: Interpretable Neural Clustering

Invariant Information Clustering for Unsupervised Image Classification and Segmentation

Deep Clustering for Unsupervised Learning of Visual Features

Representation Learning with Contrastive Predictive Coding

Structured AutoEncoders for Subspace Clustering

Neural Information Processing

Deep Adaptive Image Clustering

Improved Regularization of Convolutional Neural Networks with Cutout

Improved Deep Embedded Clustering with Local Structure Preservation

Deep Clustering via Joint Convolutional Autoencoder Embedding and Relative Entropy Minimization

Learning Discrete Representations via Information Maximizing Self-Augmented Training

Self-Taught Convolutional Neural Networks for Short Text Clustering

Deep Subspace Clustering with Sparsity Prior

A model-based approach for text clustering with outlier detection

Joint Unsupervised Learning of Deep Representations and Image Clusters

Multiple Kernel k-Means Clustering with Matrix-Induced Regularization

Deep Residual Learning for Image Recognition

Unsupervised Deep Embedding for Clustering Analysis

Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks

Skip-Thought Vectors

On comparing partitions

Robust Subspace Clustering via Thresholding Ridge Regression

Distributed Representations of Sentences and Documents

Graph Degree Linkage: Agglomerative Clustering on a Directed Graph

An Analysis of Single-Layer Networks in Unsupervised Feature Learning

Spectral Embedded Clustering: A Framework for In-Sample and Out-of-Sample Spectral Clustering

Semi-Supervised Recursive Autoencoders for Predicting Sentiment Distributions

Deconvolutional networks

Stacked Denoising Autoencoders: Learning Useful Representations in a Deep Network with a Local Denoising Criterion

Understanding the Power of Clause Learning

Spectral Curvature Clustering (SCC)

Fast unfolding of communities in large networks

The Relationships Among Various Nonnegative Matrix Factorization Methods for Clustering

Greedy Layer-Wise Training of Deep Networks

Dimensionality Reduction by Learning an Invariant Mapping

Self-Tuning Spectral Clustering

X-means: Extending K-means with Efficient Estimation of the Number of Clusters