Wednesday, November 5, 2014

Review: Thoughtful Machine Learning, by Matthew Kirk, O'Reilly Media

Unfortunately, the first edition of Thoughtful Machine Learning feels like a first edition.

The writing is nothing special, and at times, inconsistent. For example, "Neural Networks" and "Neural Nets" are used interchangeably. More importantly, some of the examples are confusing. The calculations and numbers used in the fraud detection example are not well explained.

Mistakes in the figures and tables are numerous (refer to the book's errata page). Some of the figures are overly simplistic and don't add anything. Describing a noisy party is enough, a diagram of 3 people standing next to a washing machine is overkill. To accompany the description of gradient descent, there is a figure of a person on a hill.

On the plus side, the ruby code examples for each chapter may be useful to some readers. To get them running, readers are largely on their own; only one chapter has a README.

Overall, I was disappointed. Perhaps the the second edition will be better.

No comments: