Scientific progress is predicated on quantifiable advances. The questions of what to measure and how to measure it are central to empirical computer science. From compiler to distributed machine learning algorithms, setting up experiments and reporting their outcome is something we keep getting wrong. This course will provide motivation in the form of a number of cautionary tales, and then will give simple techniques for designing experiments along with the statistical basics suited for most common benchmarking situations. We will also introduce the concepts of reproducibility and repeatability and do some hands on exercises in R.
Conference DayFri 22 JulDisplayed time zone: Amsterdam, Berlin, Bern, Rome, Stockholm, Vienna change
10:30 - 12:00
|Experimental Evaluation of Programming Systems|
Summer SchoolFile Attached