Analytics Strategist

January 27, 2009

Recommendation Algorithm and Personalization

Filed under: Datarology, Web Analytics — Tags: , , — Huayin Wang @ 6:29 pm

Recommendation algorithm is at the heart of personalization of contents!

Why? The answer lies in the growing importance and availability of data and speed of changes. 

This of course is breaking a tradition where human knowledge and insights drive designs and decisions directly – in case of personalization, many of them are incresingly mix human manual process with data and algorithm driven process and in many other cases, it can be completely data/algorithm driven.

We should really not be too surprised about this if we stop and ask, where the human knowledge and insights come from? It based on data, many different kinds of data.  When the relevant data are sufficiently available and the learning process is well understood, put human effort in between the otherwise automatable processes can only add inefficiency. 

I have just run into this interesting post about  music recommendation, a field rich with many different ways of doing personalization/recommendation.  Here it is: Four Approaches to music recommendations.

May 19, 2006

No free lunch theorem

Filed under: business strategy, Datarology, Random Thoughts, Uncategorized — Tags: , — Huayin Wang @ 2:06 pm

There are many forms of NFL theorem. I particularly like the one when applied to optimization/search algorithm. In one version, it can be stated as ” all algorithms that search for an extremum of a cost function perform exactly the same, when averaged over all possible cost functions. In particular, if algorithm A outperforms algorithm B on some cost functions, then loosely speaking there must exist exactly as many other functions where B outperforms A.” [Wolpert and Macready (1995)], see also No Free Lunch Theorem

It is a humbling experience when meditating on it, to be reminded of the importance of contextual knowledge of the problem.

Create a free website or blog at WordPress.com.