# Studies to mitigate difference between real data and simulation for jet tagging

- Institution: CERN
- Department: Machine Learning
- Taught by: Oliver Buchmuller, Arabella Martelli, Jan Kieseler, Mauro Verzetti, Markus Stoye
- Year: 2018
- Language: English
- Subjects: Computer science
- Source: [CERN](https://videos.cern.ch/record/3019261)
- Licence: [CERN](http://copyright.web.cern.ch)
- Who can watch: Tweens profiles and older (sources give no ratings; placed by Invisible Network from the source's own words)

The aim of the studies presented is to improve the performance of jet flavour tagging on real data while still exploiting a simulated dataset for the learning of the main classification task. In the presentation we explore "off the shelf" domain adaptation techniques as well as customised additions to them. The latter improves the calibration of the tagger, potentially leading to smaller systematic uncertainties. The studies are performed with simplified simulations for the case of b-jet tagging. The presentation will include first results as well as discuss pitfalls that we discovered during our research..

0:00:00 Slide 1

0:00:29 Slide 2

0:01:21 Slide 3

0:02:22 Slide 4

0:03:51 Slide 5

0:05:19 Slide 6

0:07:23 Slide 7

0:09:11 Slide 8

0:10:20 Slide 9

0:10:34 Slide 10

0:11:20 Slide 11

0:12:00 Slide 12

0:14:16 Slide 13

0:14:37 Slide 14

0:15:03 Slide 15

0:15:39 Slide 16

0:17:10 Slide 17

0:17:38 Slide 18

0:18:22 Slide 19

0:19:07 Slide 20

## Lectures

1. [Studies to mitigate difference between real data and simulation for jet tagging](https://invisiblenetwork.ai/learn/cern/3019261/2154d8b9732a), 21 min (CERN's player)