Computer Science 4442A/B Lecture Notes - Lecture 2: Feature Vector, Overfitting

• School: Western University
• Department: Computer Science
• Course: Computer Science 4442A/B
• Professor: Olga Veksler

Document Summary

Single feature classifier (length: find the best length (l) threshold, if (fish_length < l, find the best length l threshold then salmon; fish length > l fish length < l else, at l = 5, misclassified. Sea_bass; classify as salmon classify as sea bass: for example, at l = 5, misclassified, 1 sea bass, 16 salmon bass salmon. 0: classification error (total error) and so the classification error (total error) 17/50 = 34, after going through all possible thresholds, we find that the best l = 9, but there is still. 20% of fish that remain misclassified: so, how do we proceed, what do we do, we learn that length is a poor feature alone, try another feature (weight, salmon tends to be lighter. C salmon sea bass methods: feature vector [length, lightness, this gives us a graph that looks like this -> Lightness length: now fish are classified best at lightness threshold of, classification error 4%