This repository provides a PyTorch implementation of the Convolutional Sequence Embedding Recommendation Model (Caser), a model designed for personalized top-N sequential recommendation. It is intended for researchers and practitioners in the field of recommender systems who need a PyTorch-based solution for sequential recommendation tasks.

How It Works

Caser leverages convolutional neural networks to capture sequential patterns in user behavior. It embeds items into a latent space and then applies convolutional filters to extract features from sequences of these embeddings. These features are then combined with user embeddings and passed through fully-connected layers to predict the next item a user is likely to interact with. This approach allows Caser to model complex, non-linear relationships within user interaction sequences.

Quick Start & Requirements

• Install with pip install -r requirements.txt.
• Requires Python 2 or 3, PyTorch v0.4+, NumPy, and SciPy.
• Data should be formatted as train.txt and test.txt, with each line containing user item rating triplets ordered by time. Ratings are converted to 1.

Highlighted Details

• Implements the Caser model from the WSDM '18 paper.
• Claims potential for better performance than reported in the original paper, with mAP reaching 0.17 on a test set with specific configurations (d=50, L=5, T=3).
• Heavily built on the Spotlight library for utilities and interactions.

Maintenance & Community

• No explicit information on maintainers, community channels, or roadmap is provided in the README.

Licensing & Compatibility

• The README does not specify a license. Compatibility for commercial use or closed-source linking is not addressed.

Limitations & Caveats

The project relies on PyTorch v0.4+, which is an older version and may have compatibility issues with current PyTorch releases. The README does not specify a license, which could be a blocker for commercial adoption.