Backy is a neural network which is using the backpropagation algorithm. (Written in Googles Dart). Please report all errors you can find to me.
The Neuron The neuron defines how the output is computed and in what range...
It can be instanciated with any number of layer dimensions. For example: [2, 3, 1] which produces a net with 3 layers. The input layer has two inputs and the output layer has 1 output neuron. The hidden layer has 3 neurons.
Use the "train"-method to tell the net what you expect from a certain input. net.train(<input>, <expected>);
e.g. train an XOR network:
net.train([-1, -1], [ 1]); net.train([-1, 1], [-1]); net.train([ 1, -1], [-1]); net.train([ 1, 1], [ 1]);
Once the network is trained, you can use it and it will return the output:
<expected> = net.use(<input>); print(net.use([-1, 1])); // prints probably: [-.9988, .9988]
The network needs usually many trainingsteps in orderto find the right weights and therefore the solution. Use the trainer in order to train backy more comfortably.
// 1. var neuron = new TanHNeuron(); // returnes floatingpoint values between -1 and 1 var student = new Backy([2, 2, 1], neuron); var trainer = new Trainer(backy: student, maximumReapeatingCycle: 200, precision: .1); // 2. Add the pattern whcih the network should learn trainer.addTrainingCase([-1,-1], [-1]); trainer.addTrainingCase([-1, 1], [-1]); trainer.addTrainingCase([ 1,-1], [-1]); trainer.addTrainingCase([ 1, 1], [ 1]); // 3. train all the traininCases up to 300 times and be satisfied with a precision of .1 print(trainer.trainOnlineSets()); // prints number loops it took to learn all trainingcases // 4. After that you can use the neural network print(student.use([-1,-1])); print(student.use([-1, 1])); print(student.use([ 1,-1])); print(student.use([ 1, 1]));
Add this to your package's pubspec.yaml file:
dependencies: backy: ^0.2.1
You can install packages from the command line:
$ pub get
Alternatively, your editor might support
Check the docs for your editor to learn more.
Now in your Dart code, you can use:
|0.2.1||Sep 1, 2013|
|0.2.0||Sep 1, 2013|
|0.1.0||Aug 24, 2013|
Describes how popular the package is relative to other packages. [more]
Code health derived from static analysis. [more]
Reflects how tidy and up-to-date the package is. [more]
Weighted score of the above. [more]
The package version is not analyzed, because it does not support Dart 2. Until this is resolved, the package will receive a health and maintenance score of 0.
Support Dart 2 in
The SDK constraint in
pubspec.yaml doesn't allow the Dart 2.0.0 release. For information about upgrading it to be Dart 2 compatible, please see https://www.dartlang.org/dart-2#migration.
dartdoc successfully runs on your package's source files. (-10 points)
Dependencies were not resolved.