Radiate 1.6.4

.NET port of Rust crate radiate.

Radiate

Nugets.

• Radiate - Core library
• Radiate.Data - Small toy datasets for testing implementations. Radiate.Examples uses these for examples.
• Radiate.Extensions - Extensions to the core library.

Build & run examples

dotnet run --project radiate.examples --configuration Release

Overview

Radiate consists of two main concepts:

1. Data pre-processing for model input/output

Radiate input/output data operates through a Tensor object, a 3D float array wrapper which implements linear algebra functionality used throughout the library. To facilitate the pre-processing operations, Radiate uses a TensorFrame - a collection of model feature tensors & target tensors. The TensorFrame allows you to do common data pre-processing operations like Batch, Layer, Shift, Split, Reshape, Pad, Shuffle, Kernel, Transform. These options are held within the TensorFrame for prediction data processing.

1. Batch - Set a batch size to train on.
2. Layer - Layer data by n rows.
3. Split - Split the data into a training set and testing set. Default is 75% split training, 25% testing.
4. Reshape - Reshape the row vector to a shape of (height, width, depth), useful for images.
5. Pad - Pad an image Tensor with n zeros.
6. Shuffle - Shuffle the rows of the dataset randomly.
7. Kernel - Add kernel transform for the features, possible options are RBF, Polynomial, and Linear (None).
8. Transform - Transform the feature or/and target data. Options are Normalize, Standardize, OHE (One Hot Encode), and Image (divide data point by 255).

2. Model/Engine configuration

Each AI/ML algorithm comes with its own set of configuration values. See the links above on each algorithm to view specifics.

• SupervisedEngine
  1. Random Forest
  2. Support Vector Machine
  3. MultiLayer Perceptron
  4. KNearest Neighbors
• GeneticEngine
  1. Encode problems so they can be evolved through evolutionary computing
  2. Powerful Expression Tree/Graph based genetic engine for evolving complex problems.

Make Preditions

After training, the resting model can be put in a PredictionHarness<T> to make live predictions.

var (inputs, targets) = await new XOR().LoadDataSet();

var data = new TensorFrame(features, targets).Transform(Norm.Standardize);

var neuralNet = new MultilayerPerceptron()
    .Layer(new DenseInfo(64, Activation.ReLU))
    .Layer(new DenseInfo(1, Activation.Linear));

var result = neuralNet.Fit(data).Take(100).ToResult();

var model = result.GetModel();

var harness = new PredictionHarness<MultilayerPerceptron>(model, data);
var output = harness.Predict(new float[] { 0f, 1f });

var predictionResult = output.Result;       // The raw output of the model
var classification = output.Classification; // The class of the prediction
var confidence = output.Confidence;         // The regression confidence of the prediction

Random

All random numbers are generated by calling RandomRegistry.GetRandom(). This allows for random seeds to be set for model building.

// Every Random within this scope will use the same seed - resulting in the same result every time.
var (inputs, targets) = await new XOR().LoadDataSet();
RandomRegistry.Using(new Random(11), _ =>
{
    var data = new TensorFrame(features, targets).Transform(Norm.Standardize);

    var nn = new MultilayerPerceptron(grad)
        .Layer(new DenseInfo(64, Activation.ReLU))
        .Layer(new DenseInfo(1, Activation.Linear));

    var result = nn.Fit(data).Take(MaxEpochs).ToResult();
});

Examples

Datasets coming from Radiate.Data

Convolutional Neural Network on MNist handwritten digets dataset

<img src="https://machinelearningmastery.com/wp-content/uploads/2019/02/Plot-of-a-Subset-of-Images-from-the-MNIST-Dataset-1024x768.png" width="300px">

const int FeatureLimit = 5000;
const int BatchSize = 128;
const int MaxEpochs = 10;

var (rawInputs, rawLabels) = await new Mnist(FeatureLimit).GetDataSet();

var data = new TensorFrame(rawInputs, rawLabels)
   .Reshape(new Shape(28, 28, 1))
   .Transform(Norm.Image, Norm.OHE)
   .Batch(BatchSize)
   .Split();

var neuralNet = new MultilayerPerceptron()
   .Maximizing(AccuracyType.Classification)
   .Layer(new ConvInfo(32, 3))
   .Layer(new MaxPoolInfo(2))
   .Layer(new FlattenInfo())
   .Layer(new DenseInfo(64, Activation.Sigmoid))
   .Layer(new SoftmaxInfo(data.OutputCategories))

var result = neuralNet.Fit(data)
    .Take(MaxEpochs)
    .ToResult();

Evolve a string

var target = "Austin, TX";
var codex = Codecs.CharCodex(target.Length);

var engine = Engine.Genetic(codex)
    .Limit(Limit.Accuracy(Target.Length))
    .SurvivorSelector(new ElitismSelector<CharGene>())
    .ToEngine(pheno => pheno.Zip(Target).Sum(pair => pair.First == pair.Second ? 1f : 0f));

Console.WriteLine(engine.Fit()
    .Peek(Displays.Metrics)
    .ToModel());

More

See the examples for how to use the API.