https://github.com/gerdm/convex_nonconvex

A personal essay about convexity problems in machine learning
https://github.com/gerdm/convex_nonconvex

Last synced: about 1 year ago
JSON representation

A personal essay about convexity problems in machine learning

• Host: GitHub
• URL: https://github.com/gerdm/convex_nonconvex
• Owner: gerdm
• License: mit
• Created: 2018-02-01T01:56:33.000Z (over 8 years ago)
• Default Branch: master
• Last Pushed: 2018-02-01T18:20:10.000Z (over 8 years ago)
• Last Synced: 2025-02-09T16:12:46.995Z (over 1 year ago)
• Language: Jupyter Notebook
• Size: 223 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# Convexity v.s. Non-Convexity Problems

[![DOI](https://zenodo.org/badge/119768373.svg)](https://zenodo.org/badge/latestdoi/119768373)

This is a brief essay about convexity in machine learning problems. It presents a quick overview of what exactly is a convex problem and a practical comparisson between two models in machine learning to contrast convex v.s. non-convex problems.

The comparisson mentioned above makes use of the $L_2$ loss function considering a linear regression model, and a deep neural network with one hidden layer, sigmoid activation function and linear output.

This repository contains the code used to make the graphs presented as Jupyter notebooks.

![Neural Network Loss Surface](./images/nnet_loss.png)