CT475: Machine Learning & Data Mining
Topic 7: Probabilistic Machine Learning
Dr Michael Madden, National University of Ireland Galway
This course is normally delivered in a classroom setting, but I am posting some recorded lectures online.
Lecture Videos (see below for details of contents)
- Part 1: Review of Probability Basics (1:00): http://youtu.be/dr_TPEsrD8s
- Part 2: Probabilistic Classifiers (1:20): http://youtu.be/MHcHUGPKCIc
Lecture Slides (PDF):
http://datamining.it.nuigalway.ie/documents/CT475_07_ProbabilisticML.pdf
Calculations for Spam Filter Example (XLSX): http://datamining.it.nuigalway.ie/documents/NaiveBayesSpam.xlsx
Contents of Part 1: Review of Probability Basics
- Introduction & Learning Objectives
- Why Consider Uncertainty?
- Summary of Techniques for Handling Uncertainty
- Review of Probability
- Probability Notation
- Axioms of Probability
- Unconditional and Conditional Probability
- Joint Probability Distribution
- Independence & Conditional Independence
- Product Rule, Total Probability, & Bayes’ Rule
Contents of Part 2: Probabilistic Classifiers
- Reasoning with Bayes’ Rule
- Challenges in Estimating Probabilities
- Bayes’ Rule: Example
- Bayes’ Rule with Normalisation
- Bayes’ Rule: Combining Evidence & Updating
- Naïve Bayes Classifier
- Example: Play Tennis
- Example: Bayesian Spam Filter
Small Tutorials
- Excel-Based Tutorial on Computing a Decision Matrix and Plotting a ROC Curve
- Excel-Based Tutorial on Paired T-Tests for Comparing Classifier Results.