Wednesday 12:45–14:45 in Startberry, Grochowska 306/308, 2 floor

Introduction to Recommendation Systems - part 2

Piotr Bigaj, Jakub Gasiewski, Przemek Kepczynski

Audience level:
Novice

Description

This introduction aims in presentation theoretical background as well as practical exercises of basing building blocks of Recommendation Systems. After this tutorial you will be able to implement your won Recommendation System from the scratch on real data.

Abstract

The tutorial will be divided into following sessions:

  1. Introduction: motivation and goal of RS
  2. Memory-based Collaborative Filtering models
    a. Theory
    i. Item-Item approach
    ii. User-User approach
    iii. Neighbourhood and similarity selection methods
    iv. Train/Dev set splitting methods
    b. Implementation of the model on selected dataset
  3. Model-based Collaborative Filtering models
    a. Theory
    i. Different approaches to matrix factorisation (ALS, SDG,…)
    ii. Quality metrics for model-based CF
    iii. Problems of CF (Sparsity, Cold Start, Gray Sheep, Neighbourhood transivity, …
    b. Implementation of the model on selected dataset
  4. Implicit Ratings in Recommendation Systems
    a. Theory
    i. Difference in approach in explicit vs Implicit RS
    ii. Changes in the model in comparison to Model-based CF
    iii. Quality Metrics for Implicit Ratings
    b. Implementation of the model on selected dataset
  5. Summary and conclusions

What is needed: Anaconda (5.0+), TensofFlow (Single CPU is fine) Practical Exercises will be provided in Jupyter.

Subscribe to Receive PyData Updates

Subscribe