# Data-driven detector signal characterization with constrained bottleneck autoencoders

arXiv: https://arxiv.org/abs/2203.04604 · HTML: https://mlsanity.com/paper/2203.04604/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 10 of its 28 references that are in the map.
- First posted on arXiv 9 Mar 2022.

## Builds on (most cited first, 10 of 10)

- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [An Introduction to Deep Generative Modeling](https://mlsanity.com/paper/2103.05180/) (2021)
- [Variational Autoencoders for Anomalous Jet Tagging](https://mlsanity.com/paper/2007.01850/) (2020)
- [Novelty Detection Meets Collider Physics](https://mlsanity.com/paper/1807.10261/) (2018)
- [A reconfigurable neural network ASIC for detector front-end data compression at the HL-LHC](https://mlsanity.com/paper/2105.01683/) (2021)
- [Information bottleneck through variational glasses](https://mlsanity.com/paper/1912.00830/) (2019)
- [Denoising Gravitational Waves with Enhanced Deep Recurrent Denoising Auto-Encoders](https://mlsanity.com/paper/1903.03105/) (2019)
- [Deep learning based pulse shape discrimination for germanium detectors](https://mlsanity.com/paper/1903.01462/) (2019)
- [Efficient Data Compression for 3D Sparse TPC via Bicephalous Convolutional Autoencoder](https://mlsanity.com/paper/2111.05423/) (2021)
- [Prediction and compression of lattice QCD data using machine learning algorithms on quantum annealer](https://mlsanity.com/paper/2112.02120/) (2021)

## Abstract (arXiv)

A common technique in high energy physics is to characterize the response of a detector by means of models tunned to data which build parametric maps from the physical parameters of the system to the expected signal of the detector. When the underlying model is unknown it is difficult to apply this method, and often, simplifying assumptions are made introducing modeling errors. In this article, using a waveform toy model we present how deep learning in the form of constrained bottleneck autoencoders can be used to learn the underlying unknown detector response model directly from data. The results show that excellent performance results can be achieved even when the signals are significantly affected by random noise. The trained algorithm can be used simultaneously to perform estimations on the physical parameters of the model, simulate the detector response with high fidelity and to denoise detector signals.