A TensorFlow-inspired neural network library built from scratch in C# 7.3 for .NET Standard 2.0, with GPU support through cuDNN
• Code refactored to use the new C# 7.3 features
• System.Memory assembly updated to the latest 4.5.1 version
• NuGet packages updated
• Added batch normalization layers • APIs refactoring and improvements • Bug fixes
• Added computation graph network type • New INeuralNetwork and IDataset APIs • Json metadata serialization and Clone APIs fixed for cuDNN layers • General bug fixes, code and performance improvements
• New optimizers added (Adam, AdaMax, RMSProp, Momentum and more) • MNIST, CIFAR-10 and CIFAR-100 datasets added • New APIs to partition a training dataset • Added new distance-based accuracy tester • APIs refactoring and improvements • Bug fixes and speed optimizations
• Added new APIs to parse a dataset from a collection of images • Code refactoring • Minor bug fixes and improvements
• NeuralNetwork.NET is now available as a NuGet package here • Added new IDataset APIs • Added new AccuracyTester option • APIs refactoring, code improvements • Minor bug fixes
• First stable release with all the main features implemented