@@ -6,21 +6,24 @@ author: Huw Campbell <huw.campbell@gmail.com>
6maintainer: Huw Campbell <huw.campbell@gmail.com>
6maintainer: Huw Campbell <huw.campbell@gmail.com>
7copyright: (c) 2016-2017 Huw Campbell.
7copyright: (c) 2016-2017 Huw Campbell.
8synopsis: Practical Deep Learning in Haskell
8synopsis: Practical Deep Learning in Haskell
9−category: System
9+category: AI, Machine Learning
10cabal-version: >= 1.8
10cabal-version: >= 1.8
11build-type: Simple
11build-type: Simple
12description:
12description:
13 Grenade is a composable, dependently typed, practical, and fast
13 Grenade is a composable, dependently typed, practical, and fast
14− recurrent neural network library for concise and precise
14+ recurrent neural network library for precise specifications and
15− specifications of complex networks in Haskell.
15+ complex deep neural networks in Haskell.
17 Grenade provides an API for composing layers of a neural network
17 Grenade provides an API for composing layers of a neural network
18− into a sequence parallel graph in a type safe manner.
18+ into a sequence parallel graph in a type safe manner; running
19+ networks with reverse automatic differentiation to calculate their
20+ gradients; and applying gradient decent for learning.
20 Documentation and examples are available on github
22 Documentation and examples are available on github
21 <https://github.com/HuwCampbell/grenade>.
23 <https://github.com/HuwCampbell/grenade>.
23extra-source-files:
25extra-source-files:
24 cbits/im2col.h
27 cbits/im2col.h
25 cbits/im2col.c
28 cbits/im2col.c
26 cbits/gradient_decent.h
29 cbits/gradient_decent.h