svenssonjoel / coppe

ML experimentation and validation tool

Coppe pronounced Cob

Coppe is an experimental EDSL for machine learning implemented in Haskell.

Purpose and aims

1. An EDSL that gives some guarantees about "OK"-ness of neural net. To begin with that input output interfaces between layers are of compatible shape.
2. Extensible EDSL. It must be possible to add new kinds of layers without rewriting the internal representation data structure for each addition.
3. Serializable/deserializable intermediate representation of networks.

Build

Coppe uses the stack build tool. To build issue the command:

stack build

and to run an example issue:

stack run