Lecture 19 – Logistic Regression Part 2, Classification
Presented by Fernando Perez
Content by Suraj Rampure, Fernando Perez, Josh Hug, Joseph Gonzalez, Ani Adhikari
- video playlist
- code HTML
- (supplementary) video on cross-entropy loss and KL divergence
The Quick Check for this lecture is due Monday, November 16th at 11:59PM. A random one of the following Google Forms will give you an alphanumeric code once you submit; you should take this code and enter it into the “Lecture 19” question in the “Quick Check Codes” assignment on Gradescope to get credit for submitting this Quick Check.
Using thresholds to convert from predicted probabilities to classifications.
Defining several metrics of classifier performance – accuracy, precision, and recall. Confusion matrices.
Using scikit-learn to compute accuracy, precision, recall, and confusion matrices.
Exploring how threshold impacts accuracy, precision, and recall. Precision-recall curves. ROC curves. AUC.
Exploring the decision boundaries that result from a logistic regression classifier, and their relationship to the model's parameters.
Linear separability. Why we sometimes need regularization for logistic regression.
Summary. Brief introduction to multiclass classification.