SimpleML is available as a NuGet package. The source code is hosted on GitHub

Release History

Version Date Description 2018-02-20 Changed mini-batch logic in the BasicFeedForwardNeuralNetwork class so that batches can 'wrap around' to the next epoch (instead of potentially partially completing at the end of each epoch).
Changed SimpleML.Containers.CostHistoryItem to store the cost at the end of each batch rather than each epoch. 2018-01-27 Added BasicFeedForwardNeuralNetwork class. 2017-04-30 Removed interface/class Containers.Persistence.IFile and File (moved functionality to FrameworkAbstraction project).
Added classes Containers.Persistence.MnistLabelFileReader and MnistImageFileReader for future reading of MNIST digits.
Removed LogisticRegressionGradientDescentOptimizer class.
Implemented logistic regression workflow (SimpleML.Samples\Resources\Logistic Regression Regularization Comparison.xml).
Added cancellation functionality to classes / modules...
Updated the version of NMock2.dll in \Referenced Libraries folder to 2.1 (file version 2.0.50727).
Updated cost calculation in linear regression workflows to divide by the number of training examples (using class MultivariateLinearRegressionCostFunctionCalculator).
Added additional logging and metric logging functionality to various classes.
Added many missing mockery.VerifyAllExpectationsHaveBeenMet() method calls to unit tests.
Added many missing unit tests (esp for metric logging functionality). 2016-10-08 First released version.
Matrix classes, plus linear regression including both feature scaling and polynomial feature generation.
Logistic regression classes (not fully tested).


SimpleML is licensed under the Apache License, Version 2.0.