Tutorials

Beginner Tutorials

Julia & Lux for the Uninitiated

How to get started with Julia and Lux for those who have never used Julia before.

Fitting a Polynomial using MLP

Learn the Basics of Lux by fitting a Multi-Layer Perceptron to a Polynomial.

Training a Simple LSTM

Learn how to define custom layers and train an RNN on time-series data.

Use SimpleChains.jl as a Backend

Learn how to train small neural networks really fast on CPU.

Fitting with Optimization.jl

Learn how to use Optimization.jl with Lux (on GPUs).

Pre-Built Deep Learning Models

Use Boltz.jl to load pre-built DL and SciML models.

Intermediate Tutorials

MNIST Classification using Neural ODE

Train a Neural Ordinary Differential Equations to classify MNIST Images.

Bayesian Neural Networks

Figure out how to use Probabilistic Programming Frameworks like Turing with Lux.

Training a HyperNetwork

Train a hypernetwork to work on multiple datasets by predicting NN parameters.

Training a PINN

Train a PINN to solve 2D PDEs (using Nested AD).

Advanced Tutorials

Neural ODE to Model Gravitational Waveforms

Training a Neural ODE to fit simulated data of gravitational waveforms.

Optimal Control with Symbolic UDE

Train a UDE and replace a part of it with Symbolic Regression.

Larger Models

WARNING

These models are part of the Lux examples, however, these are larger model that cannot be run on CI and aren't frequently tested. If you find a bug in one of these models, please open an issue or PR to fix it.

ImageNet Classification

Train Large Image Classifiers using Lux (on Distributed GPUs).

Denoising Diffusion Implicit Model (DDIM)

Train a Diffusion Model to generate images from Gaussian noises.

ConvMixer on CIFAR-10

Train ConvMixer on CIFAR-10 to 90% accuracy within 10 minutes.

Selected 3rd Party Tutorials

WARNING

These tutorials are developed by the community and may not be up-to-date with the latest version of Lux.jl. Please refer to the official documentation for the most up-to-date information.

Please open an issue (ideally both at Lux.jl and at the downstream linked package) if any of them are non-functional and we will try to get them updated.

GPU-Accelerated Physics-Informed Neural Networks

Use Machine Learning (PINNs) to solve the Heat Equation PDE on a GPU.

Weather Forecasting with Neural ODEs

Train a neural ODEs to a multidimensional weather dataset and use it for weather forecasting.

Controlling Stochastic Differential Equations

Control the time evolution of a continuously monitored qubit described by an SDE with multiplicative scalar noise.

Medical Image Segmentation

Explore various aspects of deep learning for medical imaging and a comprehensive overview of Julia packages.

Neural PDE closures

Learn an unknown term in a PDE using convolutional neural networks and Fourier neural operators.

TIP

If you found an amazing tutorial showcasing Lux.jl online, or wrote one yourself, please open an issue or PR to add it to the list!