Message Boards

6786 Views

4 Replies

21 Total Likes

View groups...

Follow this post

Share this post:

GROUPS:

Staff Picks Data Science External Programs and Systems Graphs and Networks Wolfram Language Machine Learning Neural Networks

FNetGraph: an ergonomic alternative to NetGraph

Posted 2 years ago

Keywords: FNetGraph, NetGraph, neural networks, machine learning, Wolfram language, how to use netgraph

POSTED BY: Alec Graves

4 Replies

Sort By:

Posted 2 years ago

-- you have earned *Featured Contributor Badge* Your exceptional post has been selected for our editorial column *Staff Picks* http://wolfr.am/StaffPicks and Your Profile is now distinguished by a *Featured Contributor Badge* and is displayed on the Featured Contributor Board. Thank you!

POSTED BY: EDITORIAL BOARD

Posted 2 years ago

This is an interesting idea, thanks for suggestion, Alec. I encourage you to submit this to the Wolfram Function Repository (WFR) where it gets reviewed by internal teams. If the design will prove to be fault free and show usability, it might be considered for adoption into the built in architecture.

POSTED BY: Vitaliy Kaurov

Posted 2 years ago

This is now in the Function Repository! The new version is named RuleNetGraph, and there have been several improvements: You can use Subscript to name layers Pure functions and functions supported by ElementwiseLayer and ThreadingLayer are automagically converted into valid Layers! (ElementwiseLayer taking precedence) You can provide a list of rules as the input. ResourceFunction["RuleNetGraph"][LinearLayer[64] -> LogisticSigmoid -> (#*17 + 8 &)] The implementation is fairly similar aside from some changes to improve usability, but feel free to download the source notebook from the function repository site view the latest changes.