Discriminant Function

(Expert in Information Science) Pattern Recognition#machine Learning

• A function for evaluating “class likeness”.
• Example of a discriminant function: P(input | class) (a function that returns the probability of input belonging to a certain class).
• The Nearest Neighbor Method is also one of the discriminant functions (the distance to each point represents the “class likeness”).
• Manual method for finding the discriminant function:
  • Fit a probability distribution (such as a standard distribution) to existing data#probability
  • The graph of this distribution becomes the discriminant function.
• The goal is to automate this process (machine learning).
  • This is the basis of most class classification machine learning methods.
  • Simple SVM: Using points on the boundary (support vectors), separate the classes.