Отправили вместе с @arsenyinfo из @partially_unsupervised оглавление книги Machine Learning System Design на ревью в издательство, ревью проводит N человек по M пунктам, один из ревьюеров удивил

Is the Table of Contents appropriate for the readers described in the proposal? What should be added or deleted to reflect this reader more accurately?

Yes, the TOC is quite versatile and covers a huge range of topics all important for data science projects. However, ML is normally considered much narrower: ML normaly refers to the learning algorithm alone, so a data science project could use ML (but has not to) but data science is usually considered a much broader field than ML. Now it confuses me a little that a book about ML systems covers things like data gathering and reporting as this is exactly what separates classical ML from data science.

Does this description match the reader you would expect to be interested in this topic? Why or why not?

Not really, as data scientists and software engineers are not the same roles and it is not clear to me who would profit the most. From the TOC and the title it is clear that the book mainly addresses software engineers, but data scientists (more than ML people) are the ones who really have to come up with ML systems, so I would expect data scientists to be the main target audience.

Is the title of the book appropriate to the subject?

I feel that the book is more about data science systems than ML systems as ML for me means mainly learning algorithms, so a book about ML is normally considered to deal with learning algorithms like supervised and unsupervised algorithms, and not about pipelines and data collection, model monitoring etc. An exception is MLOps, which deals exactly with how to operate ML solutions, but I feel that the title could be broader and should not necessarily contain ML at its core.

Много думал