ritenn 0.1.0

A multilayer feedforward backpropagation neural network library
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81

# RiteNN - Another RustNN Fork

[RiteNN](https://github.com/ritedb/ritenn) provides more up to date versions of neural network library written in Rust.

## Description

RustNN is a [feedforward neural network ](http://en.wikipedia.org/wiki/Feedforward_neural_network) library. 
The library generates fully connected multi-layer artificial neural networks that are trained via [backpropagation](http://en.wikipedia.org/wiki/Backpropagation).
Networks are trained using an incremental training mode.

## Updates

- By [@PsiACE](https://github.com/PsiACE)
  - More up to date.
  - Use flexbuffers instead of json, faster and more compact.
  - Make everything happy.
- By [@Felix](https://github.com/Felix-Dommes)
  - Add L2 regularization and several activation functions to the original crate. Additionally, there are a few minor improvements.
  - Lambda can be set just like the learning rate. The activation functions for hidden and output gets set in NN::new as second and third parameter respectively.

## XOR example

This example creates a neural network with `2` nodes in the input layer,
a single hidden layer containing `3` nodes, and `1` node in the output layer.
The network is then trained on examples of the `XOR` function. All of the
methods called after `train(&examples)` are optional and are just used
to specify various options that dictate how the network should be trained.
When the `go()` method is called the network will begin training on the
given examples. See the documentation for the `NN` and `Trainer` structs
for more details.

```rust
use ritenn::{NN, HaltCondition, Activation};

// create examples of the XOR function
// the network is trained on tuples of vectors where the first vector
// is the inputs and the second vector is the expected outputs
let examples = [
    (vec![0f64, 0f64], vec![0f64]),
    (vec![0f64, 1f64], vec![1f64]),
    (vec![1f64, 0f64], vec![1f64]),
    (vec![1f64, 1f64], vec![0f64]),
];

// create a new neural network by passing a pointer to an array
// that specifies the number of layers and the number of nodes in each layer
// in this case we have an input layer with 2 nodes, one hidden layer
// with 3 nodes and the output layer has 1 node
let mut net = NN::new(&[2, 3, 1], Activation::PELU, Activation::Sigmoid);
    
// train the network on the examples of the XOR function
// all methods seen here are optional except go() which must be called to begin training
// see the documentation for the Trainer struct for more info on what each method does
net.train(&examples)
    .halt_condition( HaltCondition::Epochs(10000) )
    .log_interval( Some(100) )
    .momentum( 0.1 )
    .rate( 0.3 )
    .go();
    
// evaluate the network to see if it learned the XOR function
for &(ref inputs, ref outputs) in examples.iter() {
    let results = net.run(inputs);
    let (result, key) = (results[0].round(), outputs[0]);
    assert!(result == key);
}
```


## Credit

It is a fork of the crate [nn](https://crates.io/crates/nn/) & [Felix-Dommes/RustNN](https://github.com/Felix-Dommes/RustNN) , but some adjustments and improvements have been made to the code .

## License

This library is licensed under either of:

* MIT license [LICENSE-MIT](LICENSE-MIT) or http://opensource.org/licenses/MIT
* Apache License 2.0 [LICENSE-APACHE](LICENSE-APACHE) or https://opensource.org/licenses/Apache-2.0

at your option.