Tutorials

These tutorials introduce fundamental concepts in deep learning and their realizations in MXNet. Under the basics section, you'll find tutorials covering manipulating arrays, building networks, loading and preprocessing data, etc. Further sections introduce fundamental models for image classification, natural language processing, speech recognition, and unsupervised learning. While most tutorials are currently presented in Python, we also present a subset of tutorials using the R and Scala front ends.

Python

Basics

• Introduction to NDArray - Imperative tensor operations on CPU & GPU

• Introduction to Symbol - Neural network graphs and auto-differentiation

• Introduction to Module - MXNet's high-level interface for neural network training

• MXNet Iterators - Load data for neural network training

• Image IO - Loading and pre-processing images

• Record IO - Pack free-format data in binary files

• Custom Image IO - Write high-performance data-pipelines using mxnet.image

Computer Vision

• Classifying Handwritten Digits with Convolutional Neural Networks

• Image Segmentation - Separate out distinct objects in a photograph

• Object Detection using Faster R-CNN

• Neural Art - Transfer the style of one image onto the content the content of another image

• Large Scale Image Classification - Training with 14 million images on a single machine

Natural Language Processing

• Character-Level LSTM - Generate new text, one character at a time

• Text Classification using Convolutional Neural Networks

• NCE Loss - Speed up text classification with large output layers

Speech Recognition

• Phoneme Classification - Use LSTM recurrent nets to recognize phonemes in audio

• Baidu Warp CTC - Jointly learn predictions and alignments with CTC loss

Unsupervised Learning and Generative Modeling

• Generative Adversarial Networks

• Autoencoders - Find low dimensional representations of data

• Matrix Factorization - Discover latent factors of user preference in MovieLens data

• Recommender Systems - Build a complete recommender system with matrix factorization

R

• Neural Networks with MXNet in Five Minutes

• Classifying Handwritten Digits with Convolutional Neural Networks

• Classify Real-world Images with a Pre-trained Model

• Dogs vs. Cats Classification with Fine-tuning (source RMD)

• Character-Level Language Modeling with LSTM RNNs

Scala

• Create MXNet Scala Applications with the IntelliJ IDE

• Handwritten Digit Classification with Multilayer Perceptrons

• Character-Level Language Modeling with LSTM RNNs

Contributing Tutorials

Want to contribute an MXNet tutorial? To get started, download the tutorial template.

Other Resources

• MXNet Code Examples
• MXNet Tutorials for NVidia GTC 2016