Deep Learning with MXNet Cookbook: Discover an extensive collection of recipes for creating and implementing AI models on MXNet

An in-depth analysis into the world of deep learning using Apache MXNet for flexible and efficient research prototyping and deployment to production

Key Features

• A step-by-step tutorial towards using MXNet products to create scalable deep learning applications
• Implement tasks such as transfer learning, transformers, and more with the required speed and scalability
• Analyze the performance of models and fine-tune them for accuracy, scalability, and speed

Book Description

MXNet is an open-source deep learning framework that allows you to train and deploy neural network models and implement state-of-the-art (SOTA) architectures in NLP, CV, RL, and more. With this cookbook, you will be able to construct fast, scalable deep learning solutions using Apache MXNet.

This book will start by showing you the different versions of MXNet and what version to choose before installing your library. You will learn to start using MXNet/Gluon libraries to solve classification and regression problems and get an idea on the inner workings of these libraries. This book will also show how to use MXNet to analyze toy datasets in the areas of numerical regression, data classification, picture classification, and text classification. You'll also learn to build and train deep-learning neural network architectures from scratch, before moving on to complex concepts like transfer learning. You'll learn to construct and deploy neural network architectures including CNN, RNN, LSTMs, GANs, and integrate these models into your applications. You will also learn to analyze the performance of these models, and fine-tune them for increased accuracy, scalability, and speed.

By the end of the book, you will be able to utilize the MXNet and Gluon libraries to create and train deep learning networks using GPUs and distributed computing.

What you will learn

• Understand MXNet and Gluon libraries and their advantages
• Build and train network models from scratch using MXNet
• Apply ready-to-use pre-trained models and learn to fine-tune and apply transfer learning for increased accuracy
• Train and evaluate models using GPUs and distributed computing
• Solve modern computer vision, NLP, RL, and GANs problems using neural network techniques
• Learn how vector embeddings can be used for Recommender Systems with a Collaborative Filtering approach
• Learn about Deep Convolutional GANs architectures and how to apply them to generate photo-realistic synthetic faces

Who This Book Is For

This book is ideal for data scientists, machine learning engineers, and AWS developers who want to work with Apache MXNet for building fast, scalable deep learning solutions. The reader is expected to have a good understanding of Python programming and a working environment with Python 3.6+. A good theoretical understanding of mathematics for deep learning will be beneficial.

Table of Contents

1. Up and Running with MXNet
2. Working with MXNet and Visualizing Datasets: Gluon and DataLoader
3. Solving Regression Problems
4. Solving Classification Problems
5. Analyzing images with Computer Vision
6. Understanding text with Natural Language Processing
7. Fine-Tuning Models with Transfer Learning
8. Improving Training Performance with Distributed Training
9. Deploying Deep Learning Models in Production

Andrés Pérez Torres is a Data Scientist with a specialization in Computer Vision, Geometry, and Deep Learning SOTA models. He has a strong background in Mathematics and Physics. His professional experience includes serving as a Software Development Manager at Amazon Primer Air, leading UK Engineering activities and specializing in CV algorithm design with a full data life-cycle from data acquisition, annotation to model evaluation. He has extensive experience in Amazon training programs and Udacity MOOCs in Deep Learning and Computer Vision.