Early in Saffron’s life, a couple of guys, Bob Cagle and Dean Pfutzenreuter from Open Field software on the West Coast, understood the power of our associative memories and applied it to spam filtering.  Beyond just spam, their Electronic Learning Assistant (ELLA) was a personalized email management system for any set of folders defined by each user.  As reviewed in PC User Magazine when compared against many other solutions including collaborative filtering and a naïve Bayesian, ELLA was declared “World’s Best Spam Blocker”.  It was similarly praised by other reviews in Forbes and Fortune and by end users as “near perfect”.   As one user wrote, “[I] Have been using Ella for a few days. But, after only two mistakes so far, Ella has been 100%.  As they say, it learns as it goes. Seems to be the case.”

ELLA proves the incremental, non-parametric, nonlinear learning of associative memories.  Any user – even a cave man – can simply create a folder, show ELLA a few examples though a wizard interface, and it is off and running.  No “black art” parameter tweaking.  Learning on the fly.  Add new folders at any time or teach it new cases as spam attackers also change their tactics.  By simply correcting it when it makes a filing mistake, it keeps tracking to near perfect.  This is due to the highly nonlinear representation of the memories but also to the brilliance of Bob and Dean in how they represent email.  Bob and Dean are shining examples of our customers and partners as “creative entrepreneurs”.   Given our nonlinear engine to reason about attribute interactions, they developed the 100 email attributes that also served to make ELLA so accurate.  The power of the engine was combined with the brilliance of application design.

(more…)