Self-Supervised Learning

Constrastive Learning

Implementation Steps of SimCLR

• Data Augmentation: Create two augmented views of each image.
• Encoder Network: Use a convolutional neural network (CNN) to encode the images into feature vectors.
• Projection Head: Map the feature vectors to a lower-dimensional space where contrastive loss is applied.
• Contrastive Loss: Use the NT-Xent (Normalized Temperature-Scaled Cross Entropy) loss to bring positive pairs closer and push negative pairs apart.