Neural networks

The Neural Networks section provides packages for building, training and optimising deep learning models in Julia. The packages include basic functions for building neural networks, such as gradient computation, layer tuning, recurrent models, and support for GPU computing. The tools described in the section offer flexible options for tuning models, ranging from initialising weights and selecting activation functions to applying various loss functions and optimisation rules.

The section also presents tools for integration with Python, allowing you to use Python libraries and import models from other frameworks. The ability to work with models in ONNX format is also described, making it easy to share models between different machine learning environments. The core machine learning package in Engee supports the creation of models such as multilayer perceptrons, convolutional neural networks, and generative adversarial networks (GANs). The section includes automatic differentiation using and tools for data processing and callbacks.