Linear svm

17
Linear SVM: A Case Study, First Part ( Coding And Theory) ANANDA SWARUP DAS, TECHNICAL STAFF MEMBER, KNOWLEDGE ENGINEERING AND ENCRYPTED ANALYTICS TEAM, IBM INDIA RESEARCH LABS, NEW DELHI.

Transcript of Linear svm

Page 1: Linear svm

Linear SVM: A Case Study, First Part

( Coding And Theory)ANANDA SWARUP DAS, TECHNICAL STAFF MEMBER,

KNOWLEDGE ENGINEERING AND ENCRYPTED ANALYTICS TEAM,IBM INDIA RESEARCH LABS, NEW DELHI.

Page 2: Linear svm

Spam and HamHow I read the data: Using Pandas

How My Data Looks:

Page 3: Linear svm

Processing the Text

Number of Classes in My Data

Working with Text Data

Page 4: Linear svm

TF-IDF (Term Frequency-Inverse Document Frequency Vectors)

Output

Page 5: Linear svm

TfidfTranformer and Our Purpose

Objective:

TfidfTransformer:

Page 6: Linear svm

Hyperplane Definition (Theory for Linear SVM Begins)

Page 7: Linear svm

Role of the bias

Page 8: Linear svm

The hyperplane that we want

A Criteria for Linear SVM to work

Page 9: Linear svm

Linear SVM and the Decision Rules

Page 10: Linear svm

Orthogonality of the Weight Vector

Page 11: Linear svm
Page 12: Linear svm
Page 13: Linear svm
Page 14: Linear svm

The Concept of Margin

Page 15: Linear svm

Why Margin is Expressed in terms of the unit vector

Page 16: Linear svm

Scaling the Margin

Page 17: Linear svm