Artificial Intelligence & Machine learning 3 - Linear Classification | Stanford CS221 (Autumn 2021)

Associate Professor Percy Liang
Associate Professor of Computer Science and Statistics (courtesy)

Assistant Professor Dorsa Sadigh
Assistant Professor in the Computer Science Department & Electrical Engineering Department

To follow along with the course schedule and syllabus, visit:

0:00 Introduction
0:06 Machine learning: linear classification
0:14 Linear classification framework
2:43 An example linear classifier
6:26 Hypothesis class: which classifiers?
7:34 Loss function: how good is a classifier?
10:07 Score and margin
11:55 Zero-one loss rewritten
12:43 Optimization algorithm: how to compute best?
16:28 Digression: logistic regression
17:28 Gradient of the hinge loss
19:34 Hinge loss on training data
22:34 Gradient descent (hinge loss) in Python
26:16 Summary so far

[noorulhudasaleem]

i guess there is an error at 21:31. hinge loss on [1, -1] calculates 0.5 i believe

[sanspapyrus]

It might just be me, but there's an error around 8:54. Prof. Liang states that (1, -1) was classified incorrectly, even though it was on the correct side of the decision boundary.