How to jump start your deep learning skills using Apache MXNet
The top 5 ways to immerse yourself in deep learning and MXNet.
Door number 5 (source: PublicDomainPictures.net)
AI is taking hold of a wider audience of businesses and individuals. Deep learning-enabled capabilities such as natural language processing (NLP), computer vision, and speech recognition have become a game changer for many industries. Image and voice interaction with our computing devices is now commonplace, and this is changing how we live.
Fueling the growth of deep learning is the availability of open source frameworks and libraries that allow developers of all skill levels to build deep learning models quickly and easily. One framework making waves in the deep learning community is Apache MXNet, which reached 1.0 at the end of 2017.
Learn faster. Dig deeper. See farther.
O’Reilly has been working with AWS and a great bunch of authors to develop a deeply technical library of content for developers who want to use MXNet along with its Gluon API. However, we aren’t the only ones who’ve been putting out great content—here are some of the best resources for your deep learning needs.
1. Direct from the source: The official sites
You can’t go wrong learning from the contributors to the project. Get updates as well as straightforward getting-started material both at the official Apache MXNet site and the GitHub repo.
2. Technical resources: O’Reilly and beyond
You can check out our full list of MXNet content here—from a look at the community to Jupyter Notebook articles. We cover anomaly detection to sentiment analysis and generative models to deep matrix factorization.
Amazon Web Services also offers simple ways to get started with Apache MXNet on AWS. The AWS Deep Learning AMIs come pre-installed with popular deep learning frameworks and drivers, allowing you to get started with deep learning in minutes. Amazon SageMaker, a fully-managed service to build, train, and deploy machine learning models at any scale, comes with pre-trained algorithms, and supports MXNet and Tensorflow.
The Straight Dope, by Zack Chase Lipton, is a go-to resource that starts with a crash course on MXNet then runs through supervised learning, recurrent neural networks, and recommenders systems.
If video is more your speed, you can take a look at a couple of presentations on deep learning and generative adversarial networks from Sunil Mallya, a solution architect at AWS:
- Deep Learning with Apache MXNet and Gluon
- Introduction to Generative Adversarial Networks (GAN) with Apache MXNet
3. Projects worth checking out
For a quick jaunt into projects that use MXNet, a great place to start is Julien Simon’s post “10 Deep Learning projects based on Apache MXNet,” covering deployment with Docker and Lambda functions, face recognition, object detection, and more.
4. Community successes
Great open source projects are fueled by community, and the MXNet project is no different. These companies and individuals are using MXNet to apply deep learning to solve real-world problems:
- BorealisAI: “Standardizing a Machine Learning Framework for Applied Research—PyTorch vs. MXNet“
- Wolfram: “Apache MXNet in the Wolfram Language“
- TuSimple: “Self-driving trucks enter the fast lane using deep learning“
- GumGum: “Benchmarking Training Time for CNN-based Detectors with Apache MXNet“
- Jason Xu: “Transfer Learning with MXNet Gluon“
5. Community discussions
Of course, one of the best ways to learn is to talk to your peers who are working in the same areas and using the same tools as you are:
- Join the discussions at Discuss.mxnet.io and Discuss.gluon.ai
- Email dev@apache.mxnet.org to join the Slack channel
- See how more than 200 teams used MXNet to improve accuracy of image recognition at Kaggle
This post is a collaboration between O’Reilly and Amazon. See our statement of editorial independence.