SiaNet.CPUOnly 0.2.2

A CSharp deep learning wrapper with CNTK backend

Developing a C# wrapper to help developer easily create and train deep neural network models. I am working on enhancing the interface to load data, build model, train and predict.

Load dataset (Housing regression example)

DataFrame frame = new DataFrame();

frame.LoadFromCsv(trainFile);

var xy = frame.SplitXY(14, new[] { 1, 13 });

traintest = xy.SplitTrainTest(0.25);

Load Sample Dataset (MNIST)

Downloader.DownloadSample(SampleDataset.MNIST);

var samplePath = Downloader.GetSamplePath(SampleDataset.MNIST);

train = ImageDataGenerator.FlowFromText(samplePath.Train);

validation = ImageDataGenerator.FlowFromText(samplePath.Test);

Build Model

model = new Sequential();

model.Add(new Dense(13, 12, OptActivations.ReLU));

model.Add(new Dense(13, OptActivations.ReLU));

model.Add(new Dense(1));

Build Convolution Layers

model.Add(new Conv2D(Tuple.Create(imageDim[0], imageDim[1], imageDim[2]), 4, Tuple.Create(3, 3), Tuple.Create(2, 2), activation: OptActivations.None, weightInitializer: OptInitializers.Xavier, useBias: true, biasInitializer: OptInitializers.Ones));

model.Add(new MaxPool2D(Tuple.Create(3, 3)));

model.Add(new Conv2D(8, Tuple.Create(3, 3), Tuple.Create(2, 2), activation: OptActivations.None, weightInitializer: OptInitializers.Xavier));

model.Add(new MaxPool2D(Tuple.Create(3, 3)));

model.Add(new Dense(numClasses));

Configure Training callbacks

model.OnEpochEnd += Model_OnEpochEnd;

model.OnTrainingEnd += Model_OnTrainingEnd;

model.OnBatchEnd += Model_OnBatchEnd;

Train Model

model.Compile(OptOptimizers.Adam, OptLosses.MeanSquaredError, OptMetrics.MAE, Regulizers.RegL2(0.1)); model.Train(traintest.Train, 64, 200, traintest.Test);

API Documentation: https://deepakkumar1984.github.io/SiaNet/