BERT4Rec Sequential Recommendation

Published:

An enhanced implementation of BERT4Rec that incorporates graph-based embeddings to improve sequential recommendation accuracy. This project addresses the cold-start problem by augmenting the original BERT4Rec architecture with dynamic graph layers.

Key Features

  • BERT-based sequential recommendation
  • Graph-augmented embeddings
  • Handles cold-start scenarios effectively
  • Improved performance on sparse interaction data

Technologies

  • Transformers library
  • PyTorch
  • BERT architecture
  • Sequential modeling techniques