Download presentation
Presentation is loading. Please wait.
Published byJuliet George Modified over 8 years ago
1
Alex Howard, ETHZ – Biasing Requirements – SLAC 21 st September 2011 1 Biasing Requirements and Interfaces Alex Howard ETH, Zurich Biasing Space User Requirements Geant4 Collaboration Workshop SLAC 21 st September 2011 1.Biasing Requirements 2.GRAS Interface 3.General Purpose UI? 4.Shortcomings? 5.Discussion
2
Alex Howard, ETHZ – Biasing Requirements – SLAC 21 st September 2011 2 General Requirements Biasing not widely used, why? –Documentation? –Publicity? –Difficulty of implementation? –Understanding results? –Choices/Suitability for a given problem? How to make it easier? –User interfaces? –Smart logic – attach to mass volumes of interest, prevent stupid biasing schemes/test or validate results? –Applicability to given problems? –Documentation?
3
Alex Howard, ETHZ – Biasing Requirements – SLAC 21 st September 2011 3 Space User Requirements General purpose biasing applied at the UI level MULASSIS (electronic shielding) and GRAS (general purpose space application) are examples Engineering-style tools –Self-checking statistics/variance reduction? –Attach to mass geometry/region of interest –Auto generate interfaces (Geometrical Biasing) –UI selectable (change parameters, switch on/off) As different tools/frameworks would be better to have a common solution within Geant4
4
Alex Howard, ETHZ – Biasing Requirements – SLAC 21 st September 2011 4 Physics Based Biasing Vladimir recently introduced physics based biasing for secondaries in a region (EM/Brem splitting) Currently being tested in ref-08 Other easier interfaces/implementation? How much left to the user? Reverse Monte Carlo (Adjoint) was also introduced into Geant4 a year ago. Would be good to have a common interface? (Adjoint for example requires another geometrical source/RoI definition)
5
Alex Howard, ETHZ – Biasing Requirements – SLAC 21 st September 2011 5 Geometrical Importance Biasing Importance based biasing is implemented without touching the physics list (Advantage) Requires a connection between parallel and mass geometries –Gets complicated with user RunManager (GRAS) and UI driven geometries (care with order and implementation) Previously most uses cases were concentrated on shielding studies with the biasing of neutrons Since 2007 geometrical biasing has been possible for any particle (electrons, ’s…) Common use case is electronic component shielding with complex geometry (also GDML) Easy solution: Create parallel geometry attached to region of interest/mass geometry?
6
Alex Howard, ETHZ – Biasing Requirements – SLAC 21 st September 2011 6 GRAS Interface (1) The minimal (GRAS-based) interface has (not all fully functional) the following interactive commands: /gras/biasdet/geom_biasing –switching geometrical importance biasing off or on /gras/biasdet/geometry –Choose importance sampling geometry: cylindrical, spherical or square gras/biasdet/particle –Choose particle for biasing: e+, e-, gamma, neutron, proton gras/biasdet/spacing –Choose importance sampling spacing: auto or custom
7
Alex Howard, ETHZ – Biasing Requirements – SLAC 21 st September 2011 7 GRAS Interface (2) /gras/biasdet/length –Length of biasing geometry from start /gras/biasdet/width –Radius of biasing geometry (cylindrical), side (square) /gras/biasdet/add_interface –Add an importance sampling interface at a given point (linear distance from start of biasing) /gras/biasdet/weight –Choose weight option: auto (~2), global (set value), custom (individual values) /gras/biasdet/start –Set start of the biasing geometry /gras/biasdet/update –Update geometry. Command MUST be applied before beamOn.
8
Alex Howard, ETHZ – Biasing Requirements – SLAC 21 st September 2011 8 GRAS Interface (3) Allows the user to apply an importance geometry on top of an existing application Number of interfaces and weights are automatically assigned within the user requested envelope. Only one particle can be biased, but the selection can be for electrons, photons or neutrons. The user is responsible for coinciding the bias geometry with the mass geometry and region of interest.
9
Alex Howard, ETHZ – Biasing Requirements – SLAC 21 st September 2011 9 General Purpose Extension: A future development is foreseen whereby biasing is attached to a the mass geometry (or region) and the division and application of weights is automatic. This requires further discussion as it needs generic Geant4 helper classes, but would facilitate the implementation of geometrical biasing for the user (I think)…
10
Alex Howard, ETHZ – Biasing Requirements – SLAC 21 st September 2011 10 Shortcomings (last year’s slide) Statistical analysis/accuracy of result/convergence of result Performance improvement measure Interface – command line? Why in physics list? Sensible weight values? Protection against “dumb” users? Real Examples Validation for given use cases in terms of CPU, Variance reduction and physics performance Documentation Publicity
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.