Abstract:In recent years, contrastive learning has been a highly favored method for self-supervised representation learning, which significantly improves the unsupervised training of deep image models. Self-supervised learning is a subset of unsupervised learning in which the learning process is supervised by creating pseudolabels from the data themselves. Using supervised final adjustments after unsupervised pretraining is one way to take the most valuable information from a vast collection of unlabeled data and teach from a small number of labeled instances. This study aims firstly to compare contrastive learning with other traditional learning models; secondly to demonstrate by experimental studies the superiority of contrastive learning during classification; thirdly to fine-tune performance using pretrained models and appropriate hyperparameter selection; and finally to address the challenge of using contrastive learning techniques to produce data representations with semantic meaning that are independent of irrelevant factors like position, lighting, and background. Relying on contrastive techniques, the model efficiently captures meaningful representations by discerning similarities and differences between modified copies of the same image. The proposed strategy, involving unsupervised pretraining followed by supervised fine-tuning, improves the robustness, accuracy, and knowledge extraction of deep image models. The results show that even with a modest 5% of data labeled, the semisupervised model achieves an accuracy of 57.72%. However, the use of supervised learning with a contrastive approach and careful hyperparameter tuning increases accuracy to 85.43%. Further adjustment of the hyperparameters resulted in an excellent accuracy of 88.70%.

A Survey of Comparative Learning Methods.

A Survey on Contrastive Self-Supervised Learning

Contrastive Self-supervised Learning in Recommender Systems: A Survey

Towards Graph Contrastive Learning: A Survey and Beyond

Research on unsupervised image retrieval methods based on contrastive learning

Heterogeneous Contrastive Learning for Foundation Models and Beyond

Understand and Improve Contrastive Learning Methods for Visual Representation: A Review

Self-supervised Learning: Generative or Contrastive

Contrasting the landscape of contrastive and non-contrastive learning

Contrastive estimation reveals topic posterior information to linear models

On the Comparison between Multi-modal and Single-modal Contrastive Learning

Unsupervised Structure-Adaptive Graph Contrastive Learning.

Contrastive Learning for Multiple Models in One Supernet

Contrastive Data and Learning for Natural Language Processing

Contrast Pattern Mining: A Survey

The Power of Contrast for Feature Learning: A Theoretical Analysis

Investigating Contrastive Pair Learning's Frontiers in Supervised, Semisupervised, and Self-Supervised Learning

Semantic-Enhanced Supervised Contrastive Learning.

An Empirical Study of Graph Contrastive Learning

Contrastive learning of strong-mixing continuous-time stochastic processes

Task-oriented Contrastive Learning for Unsupervised Domain Adaptation.