Physics:Quantum data analysis/Machine Learning: Difference between revisions
Create Book IV machine learning page |
Apply continuous Quantum previous-next navigation |
||
| (6 intermediate revisions by the same user not shown) | |||
| Line 1: | Line 1: | ||
{{Short description|Machine learning in particle-physics data analysis}} | {{Quantum article nav|previous=Physics:Quantum data analysis/Principle Component Analysis|previous label=Principle Component Analysis}} | ||
| |||
| |||
{{Quantum article nav}} | |||
?{{Short description|Machine learning in particle-physics data analysis}} | |||
{{Quantum data backlink|Machine Learning}} | {{Quantum data backlink|Machine Learning}} | ||
| Line 10: | Line 14: | ||
<div style="flex:1; line-height:1.45; color:#006b45; column-count:2; column-gap:32px; column-rule:1px solid #b8d8c8;"> | <div style="flex:1; line-height:1.45; color:#006b45; column-count:2; column-gap:32px; column-rule:1px solid #b8d8c8;"> | ||
'''Machine | '''Machine Learning''' is a topic in particle-physics data analysis. Machine learning in particle-physics data analysis refers to statistical and computational methods that learn patterns from simulated or recorded collision data. It is used for classification, regression, anomaly detection, detector reconstruction, fast simulation, and analysis optimisation. Machine-learning methods are widely used when the relation between detector-level observables and physics-level quantities is high-dimensional or difficult to express with simple cuts. In high-energy physics, common uses include particle identification, jet tagging, event selection, background rejection, energy calibration, and searches for rare or unexpected signatures. A machine-learning model must be validated against independent samples and control regions. Important checks include overtraining tests, stability under systematic variations, calibration of classifier scores, and comparison between simulated and observed data in regions that are not used for the signal extraction. | ||
</div> | </div> | ||
<div style="width:300px;"> | <div style="width:300px;"> | ||
[[File: | [[File:Quantum_data_analysis_machine_learning_book4_yellow.png|thumb|280px|Machine learning represented as a compact particle-physics data analysis workflow.]] | ||
</div> | </div> | ||
Latest revision as of 22:09, 20 May 2026
?
Machine Learning is a topic in particle-physics data analysis. Machine learning in particle-physics data analysis refers to statistical and computational methods that learn patterns from simulated or recorded collision data. It is used for classification, regression, anomaly detection, detector reconstruction, fast simulation, and analysis optimisation. Machine-learning methods are widely used when the relation between detector-level observables and physics-level quantities is high-dimensional or difficult to express with simple cuts. In high-energy physics, common uses include particle identification, jet tagging, event selection, background rejection, energy calibration, and searches for rare or unexpected signatures. A machine-learning model must be validated against independent samples and control regions. Important checks include overtraining tests, stability under systematic variations, calibration of classifier scores, and comparison between simulated and observed data in regions that are not used for the signal extraction.
Overview
Machine-learning methods are widely used when the relation between detector-level observables and physics-level quantities is high-dimensional or difficult to express with simple cuts. In high-energy physics, common uses include particle identification, jet tagging, event selection, background rejection, energy calibration, and searches for rare or unexpected signatures.
Validation
A machine-learning model must be validated against independent samples and control regions. Important checks include overtraining tests, stability under systematic variations, calibration of classifier scores, and comparison between simulated and observed data in regions that are not used for the signal extraction.
Analysis role
The method is useful only when it remains interpretable enough for a physics measurement. Inputs, training samples, target definitions, preprocessing choices, and uncertainty propagation should be documented so that the result can be reproduced and compared with other analyses.[1]
See also
Table of contents (60 articles)
Index
Full contents
References
- ↑ "Review of Particle Physics". Physical Review D 110 (3): 030001. 2024. doi:10.1103/PhysRevD.110.030001.
Source attribution: Physics:Quantum data analysis/Machine Learning