Skip to content
/ nda Public

scala multi-dimensional arrays with reverse-mode autodifferentiation

Notifications You must be signed in to change notification settings

avibryant/nda

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

22 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

nda

N-Dimensional Arrays in Scala

This is a toy implementation of TensorFlow-style multidimensional arrays in Scala, including a static computation graph with automatic derivation of gradients, but with compile-time enforcement and derivation of array shapes and the TF/NumPy broadcasting rules.

As a proof of concept, it currently only supports a handful of basic arithmetic operations (+, *, -, and reduce by sum), and a dense-Array CPU-based evaluator, but is easily extensible to support more operations, GPU evaluation, etc.

See src/examples/Regression.scala for a simple demonstration.

@avibryant

About

scala multi-dimensional arrays with reverse-mode autodifferentiation

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published