Towards reproducibility of research by reuse of IT best practices

About

Reproducibility of any research gives much higher credibility both to research results and to the researchers. This is true for any kind of research including computer science, where a lot of tools and approaches have been developed to ensure reproducibility. In this talk I will focus on basic and seemingly simple principles, which sometimes look too obvious to follow, but help researchers build beautiful and reliable systems that produce consistent, measurable results. My talk will cover, among other things, the problem of embedding machine learning techniques into analysis strategy. I will also speak about the most common pitfalls in this process and how to avoid them. In addition, I will demonstrate the research environment based on the principles that I will have outlined.

About the speaker

Andrey Ustyuzhanin (36) is Head of CERN partnership program at Yandex. He is involved in the development of event indexing and event filtering services which Yandex has been providing for the LHCb experiment since 2011. Prior to this he was responsible for research and development of Yandex's advertisement platform.

Since 2002 Andrey has been teaching Computer Science at Moscow Institute of Physics and Technologies (MIPT). He was a faculty advisor for MIPT team representing Russia in Microsoft's Imagine Cup who won the prize in 2005. From 2010 Andrey has been one of judges in Imagine Cup Worldwide Finals. .

0:00:00 Slide 1

0:02:53 Slide 2

0:03:12 Slide 3

0:03:25 Slide 4

0:04:09 Slide 5

0:06:12 Slide 6

0:06:26 Slide 7

0:07:41 Slide 8