The Culture … Recommendation Engine (Two-Tower Neural CF)
Two-tower neural collaborative filtering model that scores user-post affinity for The Culture, a fashion-focused social media app. Trained on implicit feedback (views, likes, comments, saves, shares, follows, skips, reports).
Architecture
- User tower: user embedding + archetype embedding -> MLP
- Post tower: post embedding + archetype embedding -> MLP
- Score: dot product of tower outputs + bias
Metrics (validation, Top-10)
- Precision@10: 0.5455
- Recall@10: 0.9024
- NDCG@10: 0.8983
Usage
See the inference / Hugging Face call cells in the training notebook for a full working example.
Part of The Culture ML model suite.
