Download presentation
Presentation is loading. Please wait.
Published byMatilda Ellis Modified over 8 years ago
1
Introduction to Machine Learning 236756 Nir Ailon Lecture 11: Probabilistic Models
2
Most of the Course So Far: Discriminative Approach “Bayes Optimal”
3
ERM Can Sometimes Be Viewed as Discriminative Approach for a ``Made Up’’ Probabilistic Method Gaussian Density
4
Class-Conditional Density Class Prior
5
Why Not Generative Approach
6
Why Generative Approach?
7
Stats 101: Maximum Likelihood Estimator (MLE)
8
Example: MLE For Biased Coin
9
Abuse of notation! Should be density… MLE for Continuous R.V.’s
10
Naïve Bayes Approach Conditional Independence
11
Naïve Bayes Classifier (Binary Case) It’s a linear model!
12
Depends on coordinate only Depends on coordinate & label Naïve Bayes Classifier (Gaussian Case) It’s a linear model!
13
(Gaussian) Naïve Bayes vs Linear Regression
14
Bayesian Reasoning
15
Bayesian Priors vs SRM
16
Because of conditional independence Posterior Bayesian Reasoning Bayes Average Laplace Smoothing
17
Difficulties in Bayes Reasoning
18
MAP
19
Summary
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.