https://github.com/arhik/booleanad

Automatic Differentiation rules for Boolean Types and functions.
https://github.com/arhik/booleanad

automatic-differentiation boolean-logic flux julia-language machine-learning

Last synced: 7 days ago
JSON representation

Automatic Differentiation rules for Boolean Types and functions.

• Host: GitHub
• URL: https://github.com/arhik/booleanad
• Owner: arhik
• License: mit
• Created: 2020-11-11T23:58:23.000Z (about 4 years ago)
• Default Branch: main
• Last Pushed: 2020-12-10T11:47:01.000Z (about 4 years ago)
• Last Synced: 2024-12-22T22:53:35.570Z (13 days ago)
• Topics: automatic-differentiation, boolean-logic, flux, julia-language, machine-learning
• Language: Julia
• Homepage:
• Size: 21.5 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        






- Defined Automatic Differentiation rules for Boolean types.

- Meant to be used with Flux.jl

This repo makes a point that adjoints can be defined for Boolean Type.

These ruleset will help learning F$\_2$ boolean functions with help of oracle.

Warning: Extremely experimental and needs verification from experts.

Example potential application:

- HTM networks.

- Mask learning.

- Wide networks with top-k.

- Graph learning.

- Boolean Function learning.

- Electronic component wiring.