https://github.com/tismoney/hessianfree

On the New method of Hessian-free second-order optimization
https://github.com/tismoney/hessianfree

curveball hessian-free neural-networks optimizer pytorch skoltech-course

Last synced: 3 months ago
JSON representation

On the New method of Hessian-free second-order optimization

• Host: GitHub
• URL: https://github.com/tismoney/hessianfree
• Owner: Tismoney
• Created: 2019-12-15T17:31:35.000Z (about 5 years ago)
• Default Branch: master
• Last Pushed: 2020-05-31T10:49:32.000Z (over 4 years ago)
• Last Synced: 2024-10-03T17:45:10.808Z (4 months ago)
• Topics: curveball, hessian-free, neural-networks, optimizer, pytorch, skoltech-course
• Language: Jupyter Notebook
• Homepage:
• Size: 4.66 MB
• Stars: 8
• Watchers: 5
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# HessianFree

Code of the project "On the New method of Hessian-free second-order optimization".

Made initially as a project for Skoltech course "Numerical Linear Algebra 2019"

## Quick guide

```python

from befree.examples.utils import fit_model

from befree.examples.print_stats import print_stats

%matplotlib inline

configs = ['befree/config/saved_configs/mnist_mlp_adam.yaml',

           'befree/config/saved_configs/mnist_mlp_sgd.yaml',

           'befree/config/saved_configs/mnist_mlp_lbfgs.yaml']

stats = dict([fit_model(config) for config in configs])

print_stats(stats)

```

## Paper

See our technical report [here](https://drive.google.com/open?id=1qy2nMyWZ8EowgGG99iJHhauXqjlauiIr).