Neural Network (DIY)

Do-it-yourself neural network with highly customizable architecture and layers

This package enables network designers to build, simulate, and customize their own neural networks. It supports both recursive and non-recursive network architectures. Each layer is contained as a separate subclass (e.g., lconv is the convolutional layer, llstm is the long short-term memory layer), therefore enabling researchers to create their own custom layers and incorporate that into the network architecture. Two application examples (handwritten digit recognition and speech recognition) are provided.