In my last post on the highlights of 2022, I mentioned the technical books that were part of my reading list. In this post, I want to give deserved honest reviews to Abhishek Thakur’s Approaching (Almost) Any Machine Learning Problem and Sebastian et al.’s Machine Learning with PyTorch and Scikit-Learn.

Like almost every professional endeavor, the path to continuous growth in Machine Learning (ML) is subjective, personal, and evolving. Subjective and personal because your educational and professional backgrounds would determine the best path. Evolving because what you would learn next would be determined by where you are in the learning journey. Therefore, my recommendations are best suited for ‘relative’ ML beginners from application development or software engineering backgrounds. These beginners can successfully work or have worked on the ‘hello world’ of ML engineering, e.g., Iris dataset and MNIST, etc. Their development background also suggests that they are very comfortable with programming. But they want to postpone the Mathematics of ML. But if you tick these boxes, regardless, then come with me.

Abhishek Thakur’s Approaching (Almost) Any Machine Learning Problem: This 300-page book is a completely hands-on turbocharge for learning ML. It assumes basic knowledge of machine and deep learning. You learn by doing. By coding. As Abhishek remarks, “it is better enjoyed with a cup of coffee while coding away.” Despite that, the book briefly introduces terms and terminologies and explains why some choices are made and the tradeoffs to be acknowledged. It also covers NLP, computer vision, and popular tabular data. It touches on stacking and ensembling approaches and concludes with model serving and deployment. This comprehensiveness gives an all-around solid footing exactly and concise. If there is one thing I would like a revised edition of the book to address, it is to include a book index.

Sebastian et al.’s Machine Learning with PyTorch and Scikit-Learn: This 741-page book would be a natural progression from Abhishek Thakur’s book. It introduces the under-the-hood details of the ML algorithms with adequate mathematical notations that would prepare you for ML-oriented linear algebra. The 19-chapter book, asides from its extensive coverage of the ML algorithms, touches on reinforcement learning, graph neural networks, and other deep net architectures like Convolutional Neural Networks. It also provides the working mechanism of PyTorch to give an understanding of its computation graphs. Unlike Abhishek Thakur’s book, which can be completed end-to-end within a specified short period, you may want to consume Sebastian et al.’s book on a need-basis, thus, serving as a long-term reference material.

What could go next? As earlier hinted, this would be determined by your next project or task. In my case, I need more knowledge of deep learning and, at the same time, yearning for more understanding and a granular grasp of the low-level details of the ML algorithms and the underlying mathematical concepts of both ML and DL models. To address these needs, I have Jeremy and Sylvain’s Deep Learning for Coders, Philip’s Coding the Matrix, and March et al.’s Mathematics for Machine Learning on my watchlist.

Have you used any of these books? What are your opinions about them? Any book you would also like to suggest?

Kindly let me know!