nicolasavru / cl-nn

A neural network implementation in Common Lisp.

cl-nn

cl-nn is a neural network implementation in Common Lisp. Training is done with backpropagation.

Usage

Installation

1. Install quicklisp. $ represents a bash shell and * an Lisp REPL.

   * (quicklisp-quickstart:install)
   * (ql:add-to-init-file)
   * (sb-ext:exit)

2. Configure ASDF2: $ mkdir -p ~/.config/common-lisp/source-registry.conf.d/

~/.config/common-lisp/source-registry.conf.d/projects.conf should contain the following contents: (:tree (:home "Documents/lisp/"))

"Documents/lisp/" is a path relative to your home directory that ASDF will search for lisp projects in.

3. Run cl-nn: $ sbcl --load run.lisp

For the sake of your sanity, suggest running sbcl through rlwrap: $ rlwrap sbcl

Runtime

Upon startup, you will be prompted to train or run a neural network. If you choose to train one, you will be prompted to generate a new network (and specify the appropriate layer sizes) or load a neural network from a file. You will then be prompted for a trainding files, learning rate, and a number of epochs to run for.

If you choose to run a neural network, you will be prompted for a file containg a (presumably trained) neural network, a test data file, and an output file.