Title : Will we ever be able to ‘discover’ scientific theory directly from data?
Speaker: Prof. Mark Gahegan
Affiliation: The University of Auckland (Centre for eResearch and Computer Science)
Time: 2 pm Thursday, 28 September, 2017
Location: 303-257
Abstract
Of all the hype surrounding Big Data, perhaps the most intriguing aspect is the claim that we may be entering a Fourth Paradigm of science—that of being data dominated; with some authors even going so far as to herald “the end of theory”. The argument goes that—with the availability of very rich data describing a given problem—it becomes possible to 'discover' (via inductive process modeling) theoretically-based explanations, as opposed to the simpler descriptions of data artifacts that are currently produced by data mining. And of course if explanatory models can be inferred, then the need for theory-based modeling conducted by humans may diminish. In many respects, creating explanatory models is the holy grail of Artificial Intelligence research. But it is a far easier challenge to inductively learn predictive models whose results are expressed via abstract statistical or machine learning notation, rather than descriptive models expressed in terms of the theoretical understanding of a scientific domain. This talk briefly outlines four paradigms used in research, examines some of the prospects for learning theory from data and gives a couple of examples of progress to date.

Seminar list