ICML'06: ornate dinner, but a little hard to swallow

Dinner at ICML was amazing- at the Music Hall of the Carnegie Museum. The most ornate room I’ve even been in- those robber barons really knew how to party.

I love Americans and their selective memory. This is the 23rd machine learning conference and, for the first time, it is being held back where the conference series started- at CMU. So, after dinner we were treated to an American view of the history of machine learning.

Michalski took his rightful place in the sun and told the story of “AQ”: a 1983 learner that received much press since when the local expert reviewed the learned rules, he dumped his own and adopted those generated by the machine.

Then someone else gave an American view of the history of machine learning (which goes something like this:)

ebl ⇒ neuralNets ⇒ analogy ⇒ reinforcementLearning ⇒ ? ⇒ SVM and graphical models ⇒ structural learning and active learning

No where in all that history was any mention of:

• the the English like Turing and Michie

• or the Australians like Quinlan who’s empirical methods clarified a murky murky field (and whose work on entrophy-based decision tree learning was recognized at IJCAI91) or Wallace (whose 1968 publications on MDL pre-dates much of the current thinking on information theory in machine learning) or Sammut (whose thesis spawned inductive logic programming)

• or my aunt (who made great scones).