Download presentation
Presentation is loading. Please wait.
Published byDarren Floyd Modified over 9 years ago
1
Filtering and Recommendation INST 734 Module 9 Doug Oard
2
Agenda Filtering Recommender systems Classification
3
Supervised Machine Learning Model-based techniques –Hill climbing (e.g., Rocchio) –Statistical classification (e.g., SVM) –Rule induction (e.g., decision trees) –Neural networks (e.g., “deep learning”) Population-Based Techniques –Instance-based learning (e.g., kNN) –Genetic algorithms
4
Adaptive Vector-Space Filtering Make Profile Vector Compute Similarity Select and Examine (user) Assign Ratings (user) Update User Model New Documents Vector Ranking Threshold Document, Vectors Rating, Vector Vector(s) Make Document Vectors Initial Profile Features Vectors
5
Latent Semantic Indexing SVD Reduce Dimensions Make Profile Vector Make Document Vectors Compute Similarity Select and Examine (user) Assign Ratings (user) Update User Model Representative Documents Sparse Vectors Matrix New Documents Sparse Vector Dense Vector Ranking Threshold Document, Dense Vector Rating, Dense Vector Dense Vector(s) Reduce Dimensions Make Document Vectors Matrix Initial Profile Features Sparse Vectors Dense Vectors
6
Linear Separators Which of the linear separators is optimal? Original from Ray Mooney
7
Maximum Margin Classification Support Vector Machine (SVM) Implies that only support vectors matter –Other training examples are ignorable. Original from Ray Mooney
8
Soft Margin SVM ξiξi ξiξi Original from Ray Mooney
9
Non-linear SVMs Φ: x → φ(x) Original from Ray Mooney
10
Training Supervised Classifiers All learning systems share two problems –They need some “inductive bias” –They must balance adaptation with generalization Overtraining can hurt performance –Performance on training data rises and plateaus –Performance on new data rises, then falls Useful strategies –Hold out a “devtest” set to find peak on new data –Emphasize exploration early, exploration later
11
Summing Up Filtering poses some unique challenges –Adversarial behavior, new terms, throughput Behavioral signals offer unique opportunities –For both static and dynamic content Supervised classifiers learn to make decisions –Two-sided training –Threshold learning
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.