Learning Deep Learning
Theory and Practice of Neural Networks, Computer Vision, Natural Language Processing, and Transformers Using TensorFlow
By Magnus Ekman
Synopsis
Magnus Ekman, a computer architect at NVIDIA, builds deep learning from first principles in the order it historically developed. The book opens with the Rosenblatt perceptron, then works through gradient-based learning, sigmoid neurons and backpropagation, and fully connected networks for classification and regression. A chapter on frameworks and network tweaks introduces the practical techniques that make deep networks trainable, before the book moves to convolutional networks for image classification, deeper pretrained models, and recurrent networks and LSTMs for time series and text autocompletion.
The language chapters are a particular strength. Ekman covers neural language models, word2vec and GloVe embeddings, sequence-to-sequence translation, and attention leading into the Transformer, then combines vision and language in an image-captioning network. A closing medley of topics and the appendices touch on later models, including GPT and BERT. Every concept is paired with code: the printed examples use TensorFlow with Keras, and matching PyTorch notebooks are available online. Readers finish with working implementations and a coherent mental model of why modern architectures are built as they are. The book was developed with the NVIDIA Deep Learning Institute and printed in full color.
Scorecard
Editorial score
How it scored
Editorial score: 65.3 / 1005 weighted criteria
- 16.5/30
Influence30%
- 17.5/25
Authority25%
- 15.0/20
Practical value20%
- 8.3/15
Currency15%
- 8.0/10
Readability10%
- 65.3/100
Editorial score
How the rank was computed
Rank No. 389 of 500 comes from a composite of 50.1: observable signals 20.7 of 55 points plus editorial review 29.4 of 45 points.
All 11 inputs
| Input | Points |
|---|---|
| Observable signals | 20.7/55 |
| Reach (Amazon ratings volume) | 7.9/16.5 |
| Consensus across best-of lists | 5.5/13.8 |
| Awards and bestseller evidence | 0.0/8.3 |
| Reader rating, sample-adjusted | 6.9/8.3 |
| Amazon sales rank | 0.4/5.5 |
| Recency | 0.0/2.8 |
| Editorial review | 29.4/45 |
| Author authority | 7.9/11.3 |
| Influence | 7.4/13.5 |
| Practical value | 6.8/9.0 |
| Readability | 3.6/4.5 |
| Currency | 3.7/6.8 |
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 thorough, well-sequenced deep learning course from perceptrons to transformers, held back mainly by its TensorFlow-first code.
Strengths
- Logical progression from perceptron to Transformer
- Strong natural language processing coverage
- Free PyTorch notebooks complement TensorFlow text
Keep in mind
Printed code uses TensorFlow while most researchers now favor PyTorch, and the 2021 text predates instruction-tuned LLMs and modern generative AI practice.
Key takeaways
- Understanding the single perceptron and gradient descent thoroughly makes every later architecture, from CNNs to transformers, easier to reason about.
- Practical tweaks such as better activation functions, initialization, normalization and regularization are what make deep networks train reliably.
- Word embeddings such as word2vec and GloVe turn text into geometry, giving networks a meaningful numeric representation of language.
- Attention lets a model weigh every input position directly, which is why the Transformer displaced recurrent networks for language tasks.
Notable facts
- Developed in partnership with the NVIDIA Deep Learning Institute; the material is also offered as a video course, Learning Deep Learning: From Perceptron to Large Language Models. Source (opens in a new tab)
About the author
Questions
What is Learning Deep Learning about?
Who should read Learning Deep Learning?
Is Learning Deep Learning still useful in 2026?
Related books
- Deep Learning Illustrated Jon Krohn, Grant Beyleveld and Aglaé Bassens
- Neural Networks for Pattern Recognition Christopher M. Bishop
- Deep Learning John D. Kelleher
- Modern Computer Vision with PyTorch V. Kishore Ayyadevara and Yeshwanth Reddy
- Grokking Deep Learning Andrew W. Trask
- Deep Learning for Vision Systems Mohamed Elgendy
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.






