2 of 12

Machine Learning Implementation for ambiguity resolution scoring function

Project Goal: Use machine learning to implement a track scoring function for ambiguity resolution

Project Roadmap: 1) Review current implementation of ambiguity resolution in ATLAS tracking code. 2) Develop interface 3) Implement current version of ambiguity resolution 4) Identify and produce appropriate samples for validation and testing 5) Identify ML approach for scoring function 6) Train, test, repeat

Project Participants: Heather Gray, Nick Cinko, you?

3 of 12

Performance Tools

Project Goal:Develop a simple software package within the ACTS framework to be used to evaluate the physics performance of track reconstruction

Project Motivation: It’s vital to check the performance to ensure that the code we’re developing reproduces or improves upon the ATLAS performance! Developing a small package will save time and effort so that everyone doesn’t need to do this for each project

Project Roadmap:1) Code/package design 2) Produce samples for testing 3) Implementation of a basic set of histograms 4) Test out the package with other workshop participants

Project Participants: Heather Gray, you?

4 of 12

ACTS for Testbeam Reconstruction

Project Goal: Use ACTS to reconstruct tracks from hits registered in a 6-plane pixelated telescope.

Project Motivation: Great beginner application, expands user-base, ...

Project Roadmap: (1) Implement simple 6 plane geoemtry (2) Parse data testbeam data from telescope (3) Align planes (4) Fit tracks (Bonus) Account for multiple scattering

Project Participants: Timon, Simone, ...

5 of 12

Implement a DAF

Project Goal: Wrap a deterministic annealing filter around Kalman

Project Roadmap:

understand Kalman filter and understand what is already there
write DAF skeleton: same input as Kalman fitter (same interface)
write unittest (same as Kalman, but with & without additional outlier -> should give the same result)
Implement weight calculation of DAF and implement annealing of DAF
(Physics) Performance studies
(Computation) Performance studies

Project Participants: Nils Braun, Paolo Calafiura, Miha Muskinja, you?

6 of 12

Implement a CKF

Project Goal: Wrap a CKF around Kalman

Project Roadmap:

understand what is already there
brainstorm and develop interface

think about filter possibilities and how they should be interfaced
think about seed possibilities and how they should be interfaced

brainstorm and develop unittest
CKF without filters and final candidate selection
Implement first possible filters and final candidate selection
(Physics) Performance studies
(Computation) Performance studies

Project Participants: Nils Braun, Aaron Soffa, Nick Styles, Sean Conlon, you?

7 of 12

Implement Belle II geometry (VXD)

Project Goal: Belle II -> ACTS and handle non-cylindrical silicon sensors easily

Project Motivation: Necessary to benchmark ACTS against existing tracking code in Belle II. Potentially generalize to other experiments.

Project Roadmap:

understand how I built PXD Geometry with �"helpers"
understand how to built SVD Geometry
write testing/validation for it
add material effects
validate extrapolation in comparison with Geant4/Genfit2

Project Participants: Nils Braun, Pablo Goldenzweig, you?

8 of 12

Implement Belle II geometry (CDC)

Project Goal: Belle II -> ACTS and handle drift chambers in ACTS

Project Motivation: Necessary to benchmark ACTS against existing tracking code in Belle II. Potentially generalize to other experiments.

Project Roadmap:

brainstorm how to model CDC as surfaces �(make sure last extrapolation will always give a state on a plane through the wire perpendicular to the track)
see if also a different layer definition is needed (probably cylindrical is fine)
built geometry and validate (as before)
brainstorm how to handle material effects/see what ACTS can already offer
validate extrapolation in comparison with Geant4/Genfit2

Project Participants: Nils Braun, Pablo Goldenzweig, Michael Eliachevitch, you?

9 of 12

Event Time Extraction with ACTS

Project Goal: Reimplement the χ² fit-based method from Belle II

Project Roadmap:

understand Kalman filter
understand and develop another extrapolator/parameterisation �for the track time also
understand the chi^2 algorithm to "align" in time
implement it with the ACTS Kalman filter
implement a full algorithm
brainstorm and develop unittests/examples
validate (maybe with Belle II CDC)

Project Participants: Nils Braun, Lu Cao, you?

10 of 12

Electron Fitter with Bremsstrahlung

Project Goal: Study the status of the Gaussian sum filter implementation, how this connects to incorporating Bremsstrahlung information

Project Roadmap: This heavily depends on (a) what is already there (and how available it is) and (b) what your personal passion is about :-)

Project Participants: Patrick Ecker, Nils Braun, you?

11 of 12

Vertex reconstruction

Project Goal: Advance primary vertex algorithm development.� a) develop a kalman vertex fitter (bonus: with timing information)� b) develop a vertex finder algorithm (e.g. iterative finder)

Project Motivation: Vertex finding/fitting is a primary ingredient for every tracking software in HEP and it’s currently only barely started in ACTS

Project Roadmap: 1) assess additional event-data-model needs, implement them� 2) implement simple finder or fitter, depending on project

Project Participants: Simone (if someone else joins), ??

12 of 12

ACTS with GPUs

Project Goal: To define a target use case for GPUs within ACTS and provide a roadmap for doing so

Project Motivation: GPUs are a powerful tool for fast processing. Many difficulties to utilization, but

Project Roadmap:

Decide on target use case (track-seeding, clustering?) and target language (OpenCL, CUDA, VulKan)
Define metrics to assess performance
Write interface to skeleton

Project Participants: Lauren*, ??

*NB when originally proposed there was some hope that a postdoc would have started to work on this.