The field of computer sciences unfortunately does not yet have a culture that requires new discoveries to be backed by rigorous experimentation, as is the norm in physics and the biological sciences. It is very common to see CS papers with biased comparisons, incomplete scientific evidence, unreproducible results or simply incorrect statistical analysis of experimental data. I believe that most of these errors are not due to malice, but to the lack of teaching computer scientists the merits of rigorous experimental methodologies.
In this course we will review the concept of experimentation, and what is its role in Computer Science research. We will discuss how to properly plan and setup a scientific experiment, and how to correctly analyze the data generated by it. Topics include statistical techniques, experiment design procedures and the reasoning behind them.
Friday, 15:15 to 18:00
UPDATED 2016-04-14: The course repository has been changed. Please check the new link!