Cool to see ClickHouse get some interest here. Been using it for about a year now and consistently impressed with the performance on analytical queries across very large (1b+ rows) tables
Couple pain points though:
* Integration Maturity - Many tools/services either don't have integration with CH or are missing features
* User Management/Security - Have to configure users in their custom XML format, only applied at database level (no table level or row level option), and doesn't plug into SSO, LDAP, etc.
* Getting "current state" for a table - e.g. some table has users and some attributes, harder than it should be to get the "current attribute value" for all users, to do analytics on
* Log Format - very challenging to pull into log aggregation tool and get helpful information from
1st Edition of this book was excellent. Gives a solid explanation of both data mining/ml techniques and the trade-offs of choosing them.
The update to the chapter of classification was needed. Previously, the section on SVM and ANN was a subpart of a chapter, spanning no more than 10 pages, glad they added more detail there.
They also spend time in early chapters talking about preprocessing and cleaning data, something that often is glossed over.
[1] https://github.blog/2017-01-19-github-data-ready-for-you-to-...