Neos : End-to-End-Optimised Summary Statistics for High Energy Physics
(2023) 20th International Workshop on Advanced Computing and Analysis Techniques in Physics Research, ACAT 2021 2438.- Abstract
The advent of deep learning has yielded powerful tools to automatically compute gradients of computations. This is because training a neural network equates to iteratively updating its parameters using gradient descent to find the minimum of a loss function. Deep learning is then a subset of a broader paradigm; a workflow with free parameters that is end-to-end optimisable, provided one can keep track of the gradients all the way through. This work introduces neos: an example implementation following this paradigm of a fully differentiable high-energy physics workflow, capable of optimising a learnable summary statistic with respect to the expected sensitivity of an analysis. Doing this results in an optimisation process that is aware... (More)
The advent of deep learning has yielded powerful tools to automatically compute gradients of computations. This is because training a neural network equates to iteratively updating its parameters using gradient descent to find the minimum of a loss function. Deep learning is then a subset of a broader paradigm; a workflow with free parameters that is end-to-end optimisable, provided one can keep track of the gradients all the way through. This work introduces neos: an example implementation following this paradigm of a fully differentiable high-energy physics workflow, capable of optimising a learnable summary statistic with respect to the expected sensitivity of an analysis. Doing this results in an optimisation process that is aware of the modelling and treatment of systematic uncertainties.
(Less)
- author
- Simpson, Nathan LU and Heinrich, Lukas
- organization
- publishing date
- 2023
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- host publication
- Journal of Physics: Conference Series
- volume
- 2438
- article number
- 012105
- conference name
- 20th International Workshop on Advanced Computing and Analysis Techniques in Physics Research, ACAT 2021
- conference location
- Daejeon, Virtual, Korea, Republic of
- conference dates
- 2021-11-29 - 2021-12-03
- external identifiers
-
- scopus:85149727518
- DOI
- 10.1088/1742-6596/2438/1/012105
- language
- English
- LU publication?
- yes
- id
- 2388ed46-b5af-4df5-bcdf-b6ba9c08e765
- date added to LUP
- 2023-04-06 13:23:34
- date last changed
- 2023-04-06 13:23:59
@inproceedings{2388ed46-b5af-4df5-bcdf-b6ba9c08e765,
abstract = {{<p>The advent of deep learning has yielded powerful tools to automatically compute gradients of computations. This is because training a neural network equates to iteratively updating its parameters using gradient descent to find the minimum of a loss function. Deep learning is then a subset of a broader paradigm; a workflow with free parameters that is end-to-end optimisable, provided one can keep track of the gradients all the way through. This work introduces neos: an example implementation following this paradigm of a fully differentiable high-energy physics workflow, capable of optimising a learnable summary statistic with respect to the expected sensitivity of an analysis. Doing this results in an optimisation process that is aware of the modelling and treatment of systematic uncertainties.</p>}},
author = {{Simpson, Nathan and Heinrich, Lukas}},
booktitle = {{Journal of Physics: Conference Series}},
language = {{eng}},
title = {{Neos : End-to-End-Optimised Summary Statistics for High Energy Physics}},
url = {{http://dx.doi.org/10.1088/1742-6596/2438/1/012105}},
doi = {{10.1088/1742-6596/2438/1/012105}},
volume = {{2438}},
year = {{2023}},
}