★★★★★ 5
Excellent Textbook for Hands-On Learning of ML
Format: Kindle
This textbook is for the serious life-long learners of machine learning. There are at least two ways to ‘consume’ this book.
For the expert in ML, this is a textbook to study as a clear comprehensive ML overview and then to dive into sections of interest or ignorance. The concepts are grounded in code examples and are well cited (with links) to sources. Further, this textbook is appropriate if you are TensorFlow-centric and want to broaden into cutting-edge ML models/tools coded in PyTorch.
For a new learner to ML, this is a textbook to DO (not just READ) with hands-on and brain-engaged. If you realize that ML is a key life-long skill for your career, consider this textbook as part of a daily learning habit (10-30 min).
From personal experience, my advice to the new learner is as follows… First, clone the GitHub repository, setup your Python environment, and study the textbook, while working through the notebooks. Go on tangents and break the code. Do this methodically as part of your daily learning habit, but do not hesitate to jump ahead several chapters to prepare for tomorrow’s meeting. There is enough excellent material here for a full year of ML adventures.
I did a similar strategy with Raschka’s first textbook. About four years ago, I had finished Andrew Ng’s Deep Learning Specialization as a student in his first cohort. I knew the concepts well but could not do the actual application coding. I was surprised how my Python coding improved by following Raschka’s clean and elegant style. And Raschka’s code examples were meaty enough to be springboards into working applications.
Several textbook editions later, what is different about this new edition?
First, it moves you through scikit-Learn (a firm foundation) to PyTorch, instead of TensorFlow. PyTorch is a better stepping-stone, both conceptually and practically. With PyTorch, you will go further with less energy, while being able to convert your efforts into TensorFlow as needed. In addition, most of the cutting-edge ML/AI/DL research is in PyTorch. It is nice to read a recent arXiv paper, clone their repository, click on the Colab tutorial, and replicate their experiments, along with picking up a ton of new coding tricks & tips. I am excited to work through these PyTorch sections to hone my skills.
Second, there is a clear recognition of model tracking and tuning practices. This is often a gap in other ML textbooks and courses. Once you progress beyond the simple demo examples in a lecture, you realize that the real work is experiments, more experiments, and still more experiments, so that you must understand what the model architecture and hyperparameters are doing to your dataset. There is good coverage of scikit-Learn pipeline, grid search, model performance, and the like.
Third, ML/AI/DL practice is rapidly evolving. Every week new ML packages/services become available that could save much grief on your current project. What is refreshing about Raschka’s textbook series is that he constantly adding cutting-edge topics because he likes to stay current and to help us stay current. Hence, this edition contains recent ML treats as: transformers, self-supervised learning, autoencoders-to-GAN, graph neural networks, DBSCAN, t-SNE (with brief mention of UMAP), and PyTorch-Lightning.
WAS THIS REVIEW HELPFUL?YesReportShare
Reviewed in the United States on February 26, 2022
,
and
Promo Codes Available: