🧠 Slime Mold Recommender

VAE-Enhanced Collaborative Filtering with Bio-Inspired Dynamics

Revolutionary Recommendation Algorithm

Inspired by the foraging behavior of slime molds, this cutting-edge recommendation system combines Variational Autoencoders (VAE) with bio-inspired network dynamics to deliver highly accurate personalized recommendations.

0.867
RMSE Score
9,724
Movies Analyzed
610
Users Evaluated
100K+
Ratings Processed

🧬 Bio-Inspired Algorithm

Mimics the intelligent foraging patterns of Physarum polycephalum (slime mold) to discover optimal pathways through user preference networks, creating more natural and intuitive recommendations.

  • Adaptive tube growth and decay
  • Reinforcement learning dynamics
  • Network optimization through biological principles

🤖 VAE Integration

Advanced Variational Autoencoder compresses high-dimensional user-item interactions into meaningful latent representations, enabling efficient similarity computation and scalable recommendations.

  • Dimensionality reduction to 20D latent space
  • Handles sparse rating matrices
  • Scalable to large datasets

⚡ High Performance

Optimized implementation with parallel processing, smart neighbor limiting, and efficient sparse matrix operations for real-time recommendation generation.

  • Multi-threaded evaluation
  • Sparse matrix optimization
  • Sub-second recommendation generation

How It Works

1. Data Compression

The VAE learns to compress user rating patterns into a compact 20-dimensional representation, capturing the essential structure of user preferences.

2. Similarity Network

Item similarities are computed in the latent space using cosine similarity, creating a network of related content with weighted connections.

3. Slime Mold Simulation

Virtual slime mold agents explore the similarity network, with tubes growing stronger along frequently traveled paths between liked items.

4. Recommendation Generation

Items with the strongest tube connections to user's liked content are surfaced as personalized recommendations.

Quick Start

# Install dependencies pip install numpy pandas scikit-learn tensorflow scipy tqdm # Clone the repository git clone https://github.com/cuilongyin/slime-mold-recommender.git cd slime-mold-recommender # Run the example python examples/movielens_demo.py
# Basic usage from src.slime_mold_recommender import SlimeMoldRecommenderV5 # Create and train the recommender recommender = SlimeMoldRecommenderV5( latent_dim=20, decay_rate=0.1, growth_rate=0.2 ) recommender.fit(ratings_df) # Get recommendations recommendations, scores = recommender.recommend( liked_items=[1, 42, 123], top_n=5 )

Explore Further

🎮 Interactive Demo 📚 API Documentation ⭐ View on GitHub