dagnn

The goal of dagnn is to provide a minimal package for quickly constructing neural network architectures with simple syntax. Essentially you specify a directed acyclic graph between neural network layers using a list of formulas, and dagnn constructs a full torch::nn_module() that can be used in downstream deep learning workflows.

Installation

You can install the development version of dagnn from GitHub with:

# install.packages("devtools")
devtools::install_github("rdinnager/dagnn")

Example

Coming soon…

An in-the-wild example of using dagnn can be found here: https://github.com/rdinnager/bioclim_intrinsic_dimension

About the logo

Logo generated using Stable Diffusion SDXL with editing by Russell Dinnage. Prompt: “An R package logo with a dragon in the shape of an R made of a neural network visualization, network graph, monochrome background”