4.3. Linear SVMs: Separable Case
The Fisher's criterion aims at maximizing the average class distance, projected onto the direction of discrimination, while having the intraclass variance normalized. While it is a reasonable measurement, for certain critical applications, one may want to remove the influence of patterns that are far from the decision boundary, because their influence on decision accuracy may not be critical. This motivates consideration of learning models (e.g., support vector machines), which, in contrast to Fisher's discriminant, maximize the minimal margin of separation [33, 358]. This notion of separation margin can be further expanded to represent a fuzzy decision region (also called soft-decision region in the literature) ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access