# Direct Learning of Systematics-Aware Summary Statistics

- Institution: CERN
- Department: Machine Learning
- Speaker: Pablo De Castro Manzano
- Year: 2018
- Language: English
- Subjects: Computer science
- Source: [CERN](https://videos.cern.ch/record/3009887)
- 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)

Complex machine learning tools, such as deep neural networks and gradient boosting algorithms, are increasingly being used to construct powerful discriminative features for High Energy Physics analyses. These methods are typically trained with simulated or auxiliary data samples by optimising some classification or regression surrogate objective. The learned feature representations are then used to build a sample-based statistical model to perform inference (e.g. interval estimation or hypothesis testing) over a set of parameters of interest. However, the effectiveness of the mentioned approach can be reduced by the presence of known uncertainties that cause differences between training and experimental data, included in the statistical model via nuisance parameters. This work presents an end-to-end algorithm, which leverages on existing deep learning technologies but directly aims to produce inference-optimal sample-summary statistics. By including the statistical model and a differentiable approximation of the effect of nuisance parameters in the computational graph, loss functions derived form the observed Fisher information are directly optimised by stochastic gradient descent. This new technique leads to summary statistics that are aware of the known uncertainties and maximise the information that can be inferred about the parameters of interest object of a experimental measurement..

0:00:00 Slide 1

0:00:42 Slide 2

0:01:34 Slide 3

0:02:08 Slide 4

0:02:38 Slide 5

0:03:10 Slide 6

0:04:23 Slide 7

0:04:45 Slide 8

0:05:17 Slide 9

0:05:46 Slide 10

0:06:14 Slide 11

0:06:51 Slide 12

0:07:21 Slide 13

0:07:56 Slide 14

0:09:21 Slide 15

0:10:18 Slide 16

0:10:51 Slide 17

0:11:34 Slide 18

0:12:50 Slide 19

0:13:48 Slide 20

0:15:00 Slide 21

0:15:24 Slide 22

0:15:42 Slide 23

0:15:58 Slide 24

0:16:09 Slide 25

0:17:36 Slide 26

0:17:40 Slide 27

0:18:23 Slide 28

0:18:27 Slide 29

## Lectures

1. [Direct Learning of Systematics-Aware Summary Statistics](https://invisiblenetwork.ai/learn/cern/3009887/3798dad6003a), 20 min (CERN's player)