sergiosantos.info

blog and works

Book: Programming Collective Intelligence

03.Sep.2008

Programming Collective Intelligence by Toby Segaran

Ever wondered what are the algorithms used in Amazon’s recommendations, how costumer service’s bots learn to speak and answer our doubts or how spam filters categorize emails? Programming Collective Intelligence explains many of those algorithms and more. The author is Toby Segaran, a software developer who works at Metaweb, the company behind Freebase. If you never checked on Freebase I suggest you give it a spin. It’s a huge and open database of categorized information.

I always found Collective Intelligence to be an interesting subject, much like adaptive computation. Trough the book’s table of contents it’s clear that it has a nice round up on a good number of algorithms and fields. It covers classification, scaling, filtering, optimization and machine learning algorithms. It introduces support vector machines, neural networks, decision trees, Bayesian filters, Page Rank and genetic algorithms.

Besides the theoretical introductions, each algorithm is used effectively on several websites APIs like Netflix, Kayak, Yahoo! Finance and Ebay. The book covers all source code from collecting the info through the API to making all routines generic enough to enable comparison between different algorithms. The chosen programming language was Python, which being a dynamic language, helps keeping the code length small.

I really enjoyed the book, specially the way you can learn about a grand set of algorithms usable in all kinds of situation, and some still on active research. However I found all the code parts that include downloading and arranging the information a bit tedious and unnecessary, but that was the author decision on keep the book self-sufficient. After this book, if you found yourself thrilled by one of the fields presented, I suggest you buy a specific book on that particular subject to get a complete understanding.