Dynamic Poisson Factorization (PyTorch)

• Reproduces Charlin et al., 2015 with a lightweight PyTorch implementation.
• Original C++ code here
• Data: grab the TSVs from the paper’s repo and drop train.tsv, validation.tsv, test.tsv, test_users.tsv into data/.

Setup We use uv for the python environment and dependency management. Install uv then run:

uv sync

Train + evaluate

uv run python src/main.py

Prints paper metrics and writes plots to assets/.