Matt Gershoff - Eigen Man

This is the Silly Series.  Do not panic.  Shielding your eyes will not preventing from contracting a little bit of crazy.  It is already too late.

It’s time to talk to Matt Gershoff.  If you don’t want a sandwich after reading this profile, there’s something wrong with you.  Make sure you read the note at the end of the profile on the background to the image below!

Tell us who Matt Gershoff is.

‘The truth is… I am EigenMan.’

Way before becoming Eigenman, I studied resource economics in graduate school and then worked for a good chunk of the ‘90s in database marketing at a major direct advertising agency in the NYC and Paris offices.  The MadMan era ended long before I got there so no scotch at the workplace; however, Starbucks was just starting its US takeover so there was plenty of coffee (not nearly as cool, but probably for the best).   I took my leave from the agency to work at a tech startup toward the end of the first Internet Bubble.  After the burst, I found myself working with clients on web analytics projects.   A few years into it, I decided that I needed to recharge. I moved to Edinburgh, had a few drams, and picked up a masters degree in Artificial Intelligence.

At Edinburgh, I came across the intelligent learning agent framework for approaching AI and machine learning problems.  This framework provides a powerful, yet simple, way to think about lots of tasks that are part of Web Analytics – testing, targeting, adaptation, attribution, and optimization all fit neatly as attributes of the intelligent agent model.  I have started a solutions company, Conductrics , which uses this approach to make it easy(ish) for online decision automation and optimization.

If someone offered you a physical interpretation of a mathematical concept, or a sandwich, which would you choose?

Ah, you can’t fool me so easily. This is a trick question, for is not the sandwich itself mathematics made physical?  If you look, you can clearly see the sandwich throughout mathematics.  For example, the Eigendecomposition of a symmetric matrix A, is UU^-1 – which is obviously an Eigensandwich.  (See also the non standard sandwich forms such as the singular value decomposition and the Hot Dog).

What is the strangest request you’ve ever had from a customer?

‘I would like rent Shanghai Surprise’. I was working at a video store at the time, but still, there is no excuse for that.

Who is your superhero-sans-cape in the web analytics community and why?

Emer Kirrane.  No ‘why’ required.

What is your favourite chat-up line?

“Hi, I’m EigenMan’ – the rest just takes care of itself really.

Please write a short poem about web analytics.  Extra marks if it is in iambic pentameter.

Eigenman won’t do iambic pentameter but will do Haiku

Certain but unsure observes yet desires to act web analytics

What is your wish for the remainder of 2011?

Sandwiches, lots of sandwiches, for everybody.

If you were to begin an email spam campaign, what would the text of your email read?

Hmm, email spam seems a bit dated. Maybe it should be something sexier, like real-time geo social mobile app spam.  Let’s see, how about ‘Free Sandwiches, right here, right now, bring your friends.’

How many boards could the Mongols hoard if the Mongol hordes got bored?

Well the easy answer is ‘42’, isn’t it?  We all know that is a bit of a lazy answer.  A quick, cheap, and not terribly relevant shout out to the Hitchhikers Guide to the Galaxy.  Still, there is some feel-goodery from being able to reference a common touchstone that we all can appreciate.  Or do we?  Here is a little shocker, it turns out that I recently discovered that a pillar of the web analytics community has not read even a single book from the five book trilogy.  Now, EigenMan doesn’t go around calling people out, but if you want a small hint, just whip up a word cloud from recent tweets that contain ‘web analytics‘ and you will get some idea who it might be.

What is the craziest thing you’ve ever done for love or cash or both?

I try to keep cash and love separate.

***

Let me make a quick note about my creepy, Doctor Who looking photo.  Since Emer decided to dub me EigenMan, I thought it might be fun/silly to construct an image using Eigenfaces.  The image was made by taking about 30 of my photos and running a technique known as Principal Components Analysis (PCA) on them.  The image is a weighted combination of five eigenvectors associated with the largest five eigenvalues of the image covariance matrix (you can think of these eigenvectors as the dimensions with the largest variance – or most amount of information).  PCA – and the related Latent Semantic Indexing (LSI) used in natural language processing (NLP) – is a useful data reduction/feature selection method that is used in lots of data analytics problems, so you may find it interesting to check it out if you are not already familiar with it.
Eigenfaces are used in facial recognition, which is not my expertise (and to be honest, I am not really an expert in eigenvectors/values either).  I have to thank my buddy Panagiotis Sourtzinos, who is an expert over at Kingston University in London, for taking time out to help me put together my eigenfaces.

Check out the full list of interviews in the Silly Series here!