# Decorrelated Jet Substructure Tagging using Adversarial Neural Networks

- Institution: CERN
- Department: Machine Learning
- Speaker: Chase Owen Shimmin
- Year: 2017
- Language: English
- Subjects: Computer science
- Source: [CERN](https://videos.cern.ch/record/3017306)
- 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)

We describe a strategy for constructing a neural network jet substructure tagger which powerfully discriminates boosted decay signals while remaining largely uncorrelated with the jet mass. This reduces the impact of systematic uncertainties in background modeling while enhancing signal purity, resulting in improved discovery significance relative to existing taggers. The network is trained using an adversarial strategy, resulting in a tagger that learns to balance classification accuracy with decorrelation. As a benchmark scenario, we consider the case where large-radius jets originating from a boosted Z' decay are discriminated from a background of nonresonant quark and gluon jets. We show that in the presence of systematic uncertainties on the background rate, our adversarially-trained, decorrelated tagger considerably outperforms a conventionally trained neural network, despite having a slightly worse signal-background separation power. We generalize the adversarial training technique to include a parametric dependence on the signal hypothesis, training a single network that provides optimized, interpolatable decorrelated jet tagging across a continuous range of hypothetical resonance masses, after training on discrete choices of the signal mass..

0:00:00 Slide 1

0:00:22 Slide 2

0:00:34 Slide 3

0:01:00 Slide 4

0:01:15 Slide 5

0:01:44 Slide 6

0:02:48 Slide 7

0:03:40 Slide 8

0:04:24 Slide 9

0:04:56 Slide 10

0:05:25 Slide 11

0:06:08 Slide 12

0:06:28 Slide 13

0:08:00 Slide 14

0:08:51 Slide 15

0:09:28 Slide 16

0:11:08 Slide 17

0:11:46 Slide 18

0:12:24 Slide 19

0:12:34 Slide 20

0:12:58 Slide 21

0:13:17 Slide 22

0:14:35 Slide 23

0:15:20 Slide 24

0:15:53 Slide 25

0:16:28 Slide 26

0:17:23 Slide 27

0:18:29 Slide 28

0:18:54 Slide 29

0:19:41 Slide 30

0:22:27 Slide 31

0:24:08 Slide 32

0:24:44 Slide 33

0:27:06 Slide 34

0:27:12 Slide 35

0:27:22 Slide 36

0:27:28 Slide 37

0:27:32 Slide 38

0:27:36 Slide 39

0:27:40 Slide 40

0:29:22 Slide…

## Lectures

1. [Decorrelated Jet Substructure Tagging using Adversarial Neural Networks](https://invisiblenetwork.ai/learn/cern/3017306/c8ccbf04dfab), 32 min (CERN's player)