Scribd
Upload
70 views12 pages

Neural Collaborative Filtering Guide

Neural Collaborative Filtering (NCF) is a deep learning approach to collaborative filtering that models user-item interactions using neural networks and latent features. It addresses limitations of traditional methods, such as sparse data and cold-start problems, by learning complex non-linear patterns. NCF is applicable in various domains, including e-commerce, movies, and music recommendations.

Uploaded by

david
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
70 views12 pages

Neural Collaborative Filtering Guide

Neural Collaborative Filtering (NCF) is a deep learning approach to collaborative filtering that models user-item interactions using neural networks and latent features. It addresses limitations of traditional methods, such as sparse data and cold-start problems, by learning complex non-linear patterns. NCF is applicable in various domains, including e-commerce, movies, and music recommendations.

Uploaded by

david
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd

Neural

Collaborative
Filtering
(NCF)
LECTURE ON RECOMMENDER SYSTEMS USING DEEP LEARNING
PREPARED BY DR DAVID F. RAKOTOMANANJARA
Introduction to
Collaborative Filtering
 - Collaborative Filtering is a method of making
recommendations based on user-item
interactions.
 - Two main types:
 • User-User Collaborative Filtering
 • Item-Item Collaborative Filtering
 - Limitation: Struggles with sparse data and
cold-start problems.
What is Neural
Collaborative Filtering?
 - NCF is a deep learning-based approach for
collaborative filtering.
 - Combines neural networks with latent features
to model user-item interactions.
 - Learns complex non-linear patterns for better
recommendations.
Latent Features and
Embeddings
 - Latent Features: Hidden patterns representing
user preferences and item properties.
 - Embeddings: Dense vector representations of
users and items.
 - Captures semantic similarities and
interactions.
NCF Architecture
Overview
 - Input Layer: User and Item IDs
 - Embedding Layers: Convert IDs into dense
vectors
 - Concatenation Layer: Combines user and item
embeddings
 - Hidden Layers: Captures non-linear
interactions
 - Output Layer: Predicts interaction probability
(e.g., click or purchase)
NCF Architecture
Overview
Mathematical
Formulation
 - User embedding: \( E_u \) and Item
embedding: \( E_i \)
 - Interaction: \( x = [E_u, E_i] \)
 - Hidden Layers: \( h = ReLU(Wx + b) \)
 - Output Layer: \( \hat{y} = Sigmoid(h) \)
Model Training and
Evaluation
 - Loss Function: Binary Cross-Entropy for implicit
feedback
 - Optimizer: Adam or SGD
 - Evaluation Metrics:
 • Accuracy
 • Precision, Recall, F1-Score
 • AUC (Area Under the Curve)
Example using
MovieLens Dataset
 - MovieLens 100K dataset:
 • 100,000 ratings from 1,000 users on 1,700
movies.
 - Convert ratings to implicit feedback (e.g., 1 if
rating >= 4, else 0)
 - Train NCF model using PyTorch or TensorFlow.
Advantages and
Limitations
 - Advantages:
 • Captures non-linear interactions
 • Learns complex latent features
 - Limitations:
 • Requires large datasets and high
computational power
 • Can be prone to overfitting.
Applications and Use
Cases
 - E-commerce Recommendations (Amazon,
Alibaba)
 - Movie Recommendations (Netflix, Hulu)
 - Music Recommendations (Spotify, Apple
Music)
 - Personalized News Feed and Ads
Conclusion and Q&A

 - Neural Collaborative Filtering is a powerful


deep learning approach for recommendations.
 - It overcomes the limitations of traditional
collaborative filtering.
 - Questions and Discussion.

You might also like