statisticianinstilettos / recommender_engines

WIP

recommender_engines

A personal collection of code for training all kinds of recommender systems that I may want to use again. Some code is from courses, some hand written.

Contains:

• Content Filtering
• Factorization Machines
• Feature Engineering (various methods for creating feature embedding vectors)
• Reinforcement Learning (Multi-Armed Bandits)

Coming soon:

• Factorization Machines in pytorch

• Deep Factorization Machines (a Deep Hybrid Recommender system similar to the FM) in pytorch

• Feature Engineering - VAE in pytorch

• Deep Hybrid Recommender system using Vampire in pytorch

• User-based and Item-based Collaborative Filtering

  • Nearest neighbors approach
  • Matrix decomposition by gradient descent
  • Nonnegative matrix factroization
  • Probabalistic Matrix Factorization
  • Deep Collaborative Filtering with autoencoders

• Association rules

• Deep outfits. ("pairs well with")

• "Describe what you are looking" for bot - Information retrieval (cosine sim with document embeddings. take in any type of document embeddings) (This is the same as content, but it uses sentiment)

• LDA recommender system