bayesian classification

16
Bayesian Classifier Gang Tao

Upload: gang-tao

Post on 20-Feb-2017

205 views

Category:

Data & Analytics


0 download

TRANSCRIPT

Page 1: Bayesian Classification

Bayesian ClassifierGang Tao

Page 2: Bayesian Classification

Algebraic GeometryComplex Analysis factal Differential equationGeometry

Dynamical SystemCombinatorial Mathematics

StatisticsComputational mathematics

Page 3: Bayesian Classification

Bayes Theorem

Page 4: Bayesian Classification
Page 5: Bayesian Classification

Bayes Theorem

Page 6: Bayesian Classification

Diachronic Interpretation

H -> HypothesisD -> DataP(H) -> Prior ProbabilityP(H|D) -> Posterior ProbabilityP(D|H) -> LikelihoodP(D) -> Normalizing Constant

Page 7: Bayesian Classification

Bayes Theorem

Original Belief Observation+ = New Belief

Page 8: Bayesian Classification

Bayes and Occam’s Razor

Page 9: Bayesian Classification

“All Models are wrong, but some of them are better

than the others”

Page 10: Bayesian Classification

Model Complexity

Page 11: Bayesian Classification

Naive Bayes

“Naive” because it is based on independence assumptionAll the attributes are conditional independent given the class

Page 12: Bayesian Classification

Naive Bayes Classifier

Page 13: Bayesian Classification

How to build a Bayesian Classifier for prediction

Prepare Data Features Extraction

Select Distribution

Model

Calculate the Probability for each attributes

Multiply All Probabilities

Label with highest

Probability

Page 14: Bayesian Classification

Advantage VS. Disadvantage

PowerfulEfficient in Space and TimeIncremental Trainer

SimpleIndependant AssumptionProbability are not relevant

Page 15: Bayesian Classification

Application of Bayesian Classifier

Spam Email Filter

Natural Language Processing

Word Segmentation

Spell Checking

Machine Translation

Pattern Recognition

Page 16: Bayesian Classification

Thank You