Week 4: Solomonoff's 'An Inductive Inference Machine'

Discussion of Week 3:  Breiman's "Statistical Modeling: The Two Cultures"

Last week's article crystallized one of the tensions that exists in Machine Learning -- whether scientists assume that the process that produces data is unknown or stochastic. The tools that we then opt to use to solve problems will reflect our beliefs about the data generation process. While those that believe that data is generated through stochastic processes will opt for statistical methods, those that believe that data generation is essentially a black box will opt for algorithmic models.

While it is my belief that the field has increasingly adopted a mix of the two cultural approaches that Breiman discusses, I wonder how much of the academic curriculum that future data science and machine learning professionals follow reflects this shift. Readers, did you find that your classwork and professors tended to fall into one of two camps that Breiman outlines?

Week 4: Solomonoff's 'An Inductive Inference Machine'

We are going old school this week, with Solomonoff's "An Inductive Inference Machine". Regarded by some machine learning scientists as the first actual paper on machine learning, this paper discusses the use of training sets, as well as iterative solution processes. I think it's worth reading just to see how far the field has come since 1957.

Check out the article here: http://world.std.com/~rjs/indinf56.pdf .

As always, your thoughts are welcome in the comments, and any suggestions for future content will be taken into consideration. 

Finally, here's a poll that will help me tailor future content for the Machine Learning Study Group series. 

Week 4 Poll