Multi-Class Classification Methods Overview
110 likes | 470 Views
Learn about different cases in multi-class classification scenarios, including separable classes and hyperplane distinctions. Understand the requirements and conditions for using multiple Perceptrons to classify distinct patterns. Explore the concepts of separability under various conditions.
Multi-Class Classification Methods Overview
E N D
Presentation Transcript
Multi-class Classification Mu-Chun Su
Case I • Each pattern class is separable from the other classes by a single hyperplane. • M classes need M Perceptrons. • A decision can not be made if • (1) more than one perceptron have positive outputs; • (2) none of the perceptron has positiveb output.
Case II • Each pattern is separable from every other individual class by a distinct hyperplane. • M classes need M(M-1)/2 perceptrons. • If x belongs to wi then • These hyperplanes have the property that
Case III • There exist M hyperplanes with the property that if a pattern x belongs to the class wi. • If the classes are separable under case 3 condition, they are automatically separable under Case 2. • There exists no indeterminate region.