Get Up To Speed on Modern Deep Learning

Data Basics, Scaling AI Marie Merveilleux du Vignaux

Deep learning is the driving force of not only today’s everyday products and services, like digital assistants and voice-enabled TV remotes, but also tomorrow’s emerging technologies, including self-driving cars. If you need a little guidance to review the basics of deep learning and understand what it consists of, this blog post will bring you up to speed. You can also dive even deeper with this guidebook, which unpacks some of the nuances and intricacies of deep learning, including an in-depth explanation (and helpful graphics) of deep learning and how it works.

→ Download the Full Ebook: An Introduction to Deep Learning

At Dataiku, we are reading and reviewing the latest books on data science, ML, and AI so that you can decide where you want to spend your time. In case you want to take things a step further with deep learning, here's a review of a full book on the subject that goes beyond our basic guide. 

So if you want to go further, “Deep Learning Illustrated: A Visual, Interactive Guide to Artificial Intelligence,” written by John Krohn, offers accessible images and exercises to understand the secrets behind deep learning. Published in August 2019, this guide will teach you all you need to know about deep learning (and maybe even a bit more).

deep learning illustrated book cover

How Deep Are We Talking? 

Indeed, deep learning is no simple subject. A deep learning algorithm is able to learn hidden patterns from the data by itself, combine them together, and build much more efficient decision rules. Deep learning is the layering of simple algorithms, called artificial neurons, into networks several layers deep. This guidebook offers both simple and more detailed explanations. However, the more pages you turn, the more complex things get. 

This book is split into four parts. Part 1, “Introducing Deep Learning,” offers a high-level overview of what deep learning is, how it evolved, and its relationship to AI, machine learning, and reinforcement learning. This part is thus understandable and enriching for all readers, no matter the level of data science or software programming experience. Parts 3 and 4 target data scientists, along with software developers, research analysts, and other more technical profiles looking to apply deep learning in their respective fields. In these parts, you will come across theory explanations and working code run-throughs from several deep learning approaches and applications (including but not limited to machine vision, natural language processing, and image generation). 

As the majority of the book is written for technical profiles, I recommend this guide to data scientists looking to further their skills or read and interact with a fresh new guide.

What’s So Special About This Guide? 

Data scientists will not only benefit from the extra reading tips offered throughout this guide, but also from in-text bit.lys that help demonstrate concepts. So as you read on, make sure to have a phone or computer near you! You will be able to play with features and data to see how the end result changes and test out some exercises.

For example, when discussing neural networks in biological and machine vision, you will be guided to the TensorFlow Playground where you will train a deep learning model to distinguish orange dots from blue dots based solely on their location on a grid. Each hidden layer includes artificial neurons specialized in recognizing one feature or property. As the network trains, you can observe how the artificial neurons increase in complexity and abstraction the deeper down in the layers they are positioned. Information is passed from layer to layer to construct the complex final shape, enabling the network to accurately predict the color of the dot. 

tensorflow playground

In that same chapter, you will also be sent to Google’s “Quick, Draw!” neural network doodling recognition program with a simple link. This AI experiment asks you to draw a picture of an object or idea and then uses a neural network to guess what your drawing represents, basing itself off of past submissions. The AI learns from each submission, thus increasing its ability to make correct predictions in the future. 

quick, draw

These are only two of the many fascinating experiments I encourage you to discover within this guide.

Go From Theory to Practice

After you read the book, you'll be ready to create your own deep learning project, and what better place to do that than in Dataiku? You can embrace the power of AI with multiple deep learning features.

  • Visual ML With a Side of Code: Dataiku provides a visual ML tool (but will also require a little bit of coding skill to define the deep learning architecture using the Keras and TensorFlow libraries.) You write the code that defines the architecture of your deep learning model and Dataiku handles the rest! 
  • ML-Assisted Labeling: Dataiku makes data labeling easy with the ML-assisted labeling plugin that helps you easily label your tabular, sound, or image datasets. 
  • Transfer Learning: The Dataiku deep learning image plugin provides pre-trained models leveraging Keras and TensorFlow to score images and obtain classes, or for feature extraction. 

You May Also Like

AI-Ready Architecture in the Cloud

Read More

Is All AutoML Created Equal?

Read More

Moving Toward a Citizen Data Science Model

Read More

How to Perform Basic ML Serving With Python, Docker, and Kubernetes

Read More