Implementation of Contracting Newton method for minimizing logistic regression model.
To plot the graphs from the paper "Convex optimization based on global lower second-order models" by N. Doikov and Yu. Nesterov (https://arxiv.org/abs/2006.08518) do the following.
-
Compile the C++ code, using the command: $ g++ main.cc optimize.cc -O2 -std=c++17 -o main
-
Download the files: "data/w8a.txt", "data/covtype_bin_sc" from https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/.
-
To run the basic (exact) methods, use $ python3 demo_basic.py
-
To run the stochastic methods, use $ python3 demo_stochastic.py
The graphs will be placed into "output/*".