No. 365 of 500 No. 34 in Deep Learning

Deep Learning

By John D. Kelleher

Print length
296 pages
Publisher
The MIT Press
Reading level
Beginner
ISBN-13
978-0262537551
Published
September 10, 2019
Written for
Non-specialists and students who want a conceptual grounding in neural networks before tackling code.

Synopsis

John D. Kelleher's entry in the MIT Press Essential Knowledge series sets out to explain deep learning, the family of techniques behind modern speech recognition, machine translation and image analysis, without assuming a technical background. He begins with the conceptual groundwork of what it means for a machine to learn a function from data, then builds up the artificial neuron, shows how neurons combine into layered networks, and traces the field's uneven history from early perceptrons through periods of neglect to the breakthroughs of the 2010s. Later chapters describe convolutional networks for images and recurrent networks, including long short-term memory units, for sequences, and they touch on autoencoders, generative adversarial networks and capsule networks.

The book's most valuable section explains how networks are trained. Kelleher walks through gradient descent and backpropagation step by step, introducing the required mathematics as he goes rather than assuming it, so a reader finishes with a working mental model of how error signals adjust a network's weights. A glossary and a closing chapter on trends and open problems round out the volume. The result suits managers, students and curious readers who want to understand what engineers mean by training a model; anyone hoping to build systems will need a hands-on companion.

Scorecard

Editorial score

How it scored

Editorial score: 62.8 / 1005 weighted criteria

  • Influence30%

    18.0/30
  • Authority25%

    18.8/25
  • Practical value20%

    11.0/20
  • Currency15%

    7.5/15
  • Readability10%

    7.5/10
  • Editorial score

    62.8/100

How the rank was computed

Rank No. 365 of 500 comes from a composite of 51.0: observable signals 22.7 of 55 points plus editorial review 28.3 of 45 points.

All 11 inputs
InputPoints
Observable signals22.7/55
Reach (Amazon ratings volume)9.9/16.5
Consensus across best-of lists5.5/13.8
Awards and bestseller evidence0.0/8.3
Reader rating, sample-adjusted6.1/8.3
Amazon sales rank1.2/5.5
Recency0.0/2.8
Editorial review28.3/45
Author authority8.4/11.3
Influence8.1/13.5
Practical value5.0/9.0
Readability3.4/4.5
Currency3.4/6.8

Each criterion is scored from 1 to 10 and weighted, and the points add up to the editorial score. The rank combines observable signals (55 percent) with this score (45 percent). Read the methodology.

Verdict

A clear, compact primer that explains how neural networks learn, best suited to readers who want concepts rather than code.

Strengths

  • Patient, stepwise explanation of backpropagation
  • Concise history of neural network research
  • Written by an active AI researcher

Keep in mind

Written in 2019, before the ChatGPT era, it says little about how deep learning scaled into today's large language models and generative AI.

Key takeaways

  1. Deep learning is a way of learning functions from data: networks map inputs to outputs by adjusting many weighted connections between simple units.
  2. Depth matters because successive layers learn increasingly abstract features, which lets networks handle raw images, audio and text without hand-crafted rules.
  3. Gradient descent and backpropagation are the two algorithms that make training possible, assigning blame for errors to each weight and nudging it accordingly.
  4. Convolutional networks suit spatial data such as images, while recurrent networks and LSTMs were designed to process sequences such as speech and text.

About the author

John D. Kelleher

John D. Kelleher holds the Chair of Computer Science at Trinity College Dublin and directs the ADAPT research center, roles he took up in 2024. He was previously a professor at Maynooth University's Hamilton Institute and an academic leader at Technological University Dublin. His research spans machine learning and natural language processing, and he co-wrote Fundamentals of Machine Learning for Predictive Data Analytics and Data Science.

Questions

Is Deep Learning by John D. Kelleher worth reading in 2026?
Yes, for readers who want the underlying ideas. Its explanations of neurons, layers, gradient descent and backpropagation remain accurate because those foundations have not changed. It was published in 2019, so it says little about the large language models and generative AI tools that arrived afterward. Pair it with a newer book on generative AI for the current landscape.
Do I need a math or programming background to read Kelleher's Deep Learning?
No programming is required, since the book contains explanations rather than code. Some comfort with basic algebra helps in the later chapters, where Kelleher works through how gradient descent and backpropagation update a network's weights. Readers without that background can still follow the main argument, and the glossary defines the technical terms used throughout the book.
What does Kelleher's Deep Learning cover?
The book covers what machine learning is, how artificial neurons combine into networks, a history of neural network research, convolutional and recurrent architectures, and the training algorithms of gradient descent and backpropagation. It also introduces autoencoders, generative adversarial networks and capsule networks, and closes with a chapter on the field's likely future directions and challenges.

Sources

As an Amazon Associate, Best AI Books earns from qualifying purchases. Best AI Books is sponsored by Iternal.ai, whose founder and CEO, John Byron Hanby IV, wrote The AI Strategy Blueprint, our #1 ranked book. How we rank books.