NicolasAG / ML-assignment1

Logistic Regression: primal and dual implementation with polynomial kernel

ML-assignment1

Logistic Regression: primal and dual implementation with polynomial kernel

Data

• x matrix in hw1x.dat
• y vector in hw1y.dat

Usage

python main.py <FLAGS> <-- will work for all questions and plot all graphs.

FLAGS:

• --normalize yes/no : will normalize or not the X matrix (default to yes)
• --use_sgd yes/no : yes=use sklearn.linear_model.SGDClassifier no=sklearn.linear_model.LogisticRegression (default to yes)
• --n_iter 10000 : number of iterations for Gradient Descent, or max number of iterations for Logistic Regression (default 10000)
• --q1d only produce plots for Q1.d) ie: logistic regression with L2 regularization
• --q1f only produce plots for Q1.f) ie: 5 x logistic regression on data applied to 5 gaussian basis functions
• --q1g only produce plots for Q1.g) ie: logistic regression on data applied to 25 gaussian basis functions with L2 regularization
• --q2 only produce plots for Q2.c) ie: polynomial kernelized logistic regression.