Machine Learning with Swift

BrainCore

This Swift library provides feedforward and recurrent neural networks. Several types of layers are present, including the inner product layer, the linear rectifier (ReLU) layer, the sigmoid layer, the RNN and LSTM layer and the L2 loss layer. BrainCore uses Metal acceleration for both training and inference stages. The definition of a new neural network looks very clear because of some pleasant syntactic sugar:

let net = Net.build { 
    [dataLayer1, dataLayer2] => lstmLayer  
    lstmLayer =>> ipLayer1 => reluLayer1 => sinkLayer1 
    lstmLayer =>> ipLayer2 => reluLayer2 => sinkLayer2  
} 

The library can be used on both iOS and macOS. It uses the Upsurge mathematical library as a dependency. It is available via CocoaPods or Carthage.

GitHub repository:  ...