2023:groups:bsmtools:mlreinterpretation
Differences
This shows you the differences between two versions of the page.
| Both sides previous revision Previous revision Next revision | Previous revision | ||
|
2023:groups:bsmtools:mlreinterpretation [2023/06/27 12:12] gregor.kasieczka [Surrogate models for object tagging] |
2023:groups:bsmtools:mlreinterpretation [2023/06/28 14:47] (current) sezen.sekmen |
||
|---|---|---|---|
| Line 1: | Line 1: | ||
| ====== ML reinterpretation ====== | ====== ML reinterpretation ====== | ||
| + | |||
| + | ==== 26 June ==== | ||
| {{:2023:groups:bsmtools:mlreint26jun.png?400|}} | {{:2023:groups:bsmtools:mlreint26jun.png?400|}} | ||
| - | ==== Full analysis recasting ==== | + | ==== 27 June ==== |
| + | |||
| + | {{:2023:groups:bsmtools:mlreint27jun_1.jpg?400|}} | ||
| + | |||
| + | {{:2023:groups:bsmtools:mlreint27jun_2.jpg?400|}} | ||
| + | |||
| + | ===== Full analysis recasting ===== | ||
| Standards for sharing models: | Standards for sharing models: | ||
| Line 11: | Line 19: | ||
| Analyses that have provided ML models: | Analyses that have provided ML models: | ||
| - | * ... | + | * [[https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/PAPERS/SUSY-2019-04/|ATLAS-SUSY-2019-04]] |
| + | * [[https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/PAPERS/SUSY-2018-30/|ATLAS-SUSY-2018-30]] | ||
| + | * ? | ||
| Discussion during the Dec '22 reinterpretation workshop: [[https://indico.cern.ch/event/1197680/timetable/?view=standard#b-485872-experience-and-feedba|link]]. | Discussion during the Dec '22 reinterpretation workshop: [[https://indico.cern.ch/event/1197680/timetable/?view=standard#b-485872-experience-and-feedba|link]]. | ||
| - | ==== Surrogate models for object tagging ==== | + | [[https://www.overleaf.com/8811915719zfjtnygcdgpv|Overleaf document for writeup]] |
| + | ===== Surrogate models for object tagging ===== | ||
| - | Surrogate model studies | ||
| Propose to build a surrogate model using the JetClass dataset [1], trying to approximate the output of a state-of-the-art attention based tagger (ParT, [1]) -- which uses low-level inputs including vertex information -- with a network only using high-level kinematics / n-subjettiness. | Propose to build a surrogate model using the JetClass dataset [1], trying to approximate the output of a state-of-the-art attention based tagger (ParT, [1]) -- which uses low-level inputs including vertex information -- with a network only using high-level kinematics / n-subjettiness. | ||
| Hamburg is preparing a simplified dataset (dropping low level features, adding ParT output, restricting to hadronic top vs light quark/gluon; reducing examples/class to 2M train / class; 500k test/class; 1M val/class) | Hamburg is preparing a simplified dataset (dropping low level features, adding ParT output, restricting to hadronic top vs light quark/gluon; reducing examples/class to 2M train / class; 500k test/class; 1M val/class) | ||
| + | |||
| + | |||
| Based on this we can test different surrogate models, Bayesian NN, explicit sampling. | Based on this we can test different surrogate models, Bayesian NN, explicit sampling. | ||
| + | |||
| [1] Paper that introduced jetClass data: [[https://arxiv.org/abs/2202.03772 | arXiv:2202.03772]] | [1] Paper that introduced jetClass data: [[https://arxiv.org/abs/2202.03772 | arXiv:2202.03772]] | ||
2023/groups/bsmtools/mlreinterpretation.1687860747.txt.gz · Last modified: 2023/06/27 12:12 by gregor.kasieczka
Page Tools
Except where otherwise noted, content on this wiki is licensed under the following license: Public Domain
