huwcampbell.com
馃挘 Machine learning which might blow up in your face 馃挘
1module Grenade (
2 -- | This is an empty module which simply re-exports public definitions
3 -- for machine learning with Grenade.
4
5 -- * Exported modules
6 --
7 -- | The core types and runners for Grenade.
8 module Grenade.Core
9
10 -- | The neural network layer zoo
11 , module Grenade.Layers
12
13
14 -- * Overview of the library
15 -- $library
16
17 -- * Example usage
18 -- $example
19
20 ) where
21
22import Grenade.Core
23import Grenade.Layers
24
25{- $library
26Grenade is a purely functional deep learning library.
27
28It provides an expressive type level API for the construction
29of complex neural network architectures. Backing this API is and
30implementation written using BLAS and LAPACK, mostly provided by
31the hmatrix library.
32
33-}
34
35{- $example
36A few examples are provided at https://github.com/HuwCampbell/grenade
37under the examples folder.
38
39The starting place is to write your neural network type and a
40function to create a random layer of that type. The following
41is a simple example which runs a logistic regression.
42
43> type MyNet = Network '[ FullyConnected 10 1, Logit ] '[ 'D1 10, 'D1 1, 'D1 1 ]
44>
45> randomMyNet :: MonadRandom MyNet
46> randomMyNet = randomNetwork
47
48The function `randomMyNet` witnesses the `CreatableNetwork`
49constraint of the neural network, and in doing so, ensures the network
50can be built, and hence, that the architecture is sound.
51-}
52
53