mullman99 - 6:50 pm on Jan 8, 2012 (gmt 0)

If you look at what Panda is doing - analyzing many more factors, each to a 'deeper' level of analysis - you start to realize what a staggering technical feat this is, even for Google. Even with an estimated few-hundred-thousand to one million servers and the largest fiber network, they can only do this analysis by taking a moment-in-time 'snapshot' of their database and performing these calculations.

That's my understanding of why it's run periodically - that, and so it can re-calibrate itself, that 'machine learning' process. To my mind, it's only a matter of time - optimizing the software, optimizing the network to run it - until it's run alongside, or integrated into the 'regular' algo.

Michael