Description

Code to conduct experiments for the paper Regularization and acceleration of Gauss-Newton method.

Overview

• run_experiments.py — main script to perform experiments;
• oracles.py — contains classes for optimization criteria;
• opt_utils.py — auxiliary functions for optimizers;
• optimizers.py — contains Gauss-Newton optimization algorithms;
• benchmark_utils.py — routines for designed experiments;
• plotting.py — routines for plotting results;
• print_time.py — routines for registering time measurements.

Print help in command line in repository directory to list all hyperparameters of the experiments:

    python run_experiments.py -h

Run the following command in command line in repository directory to obtain all experiment data in current directory:

    python run_experiments.py

Requirements

• NumPy;
• Matplotlib;
• Seaborn.

References

[1] Yudin N.E., Gasnikov A.V. Regularization and acceleration of Gauss-Newton method // Computer Research and Modeling, 2024, vol. 16, no. 7, pp. 1829-1840, doi: https://doi.org/10.20537/2076-7633-2024-16-7-1829-1840.