SlideShare una empresa de Scribd logo
1 de 66
Descargar para leer sin conexión
Optimisation of
           Multi-Object Spectroscopy
                  in Astronomy
                      Brent Miszalski
                     SALT Research Fellow
                       brent@saao.ac.za



Sunday 18 March 12
Overview

 • Galaxy redshift surveys
 • Multi-object spectroscopy (MOS)
 • MOS field configuration by
        simulated annealing
 • MOS at the Southern African Large
        Telescope (SALT)

    Miszalski et al. 2006, MNRAS, 371,1537
Sunday 18 March 12
NGC 1376
Sunday 18 March 12
M 101
Sunday 18 March 12
Hubble Ultra Deep Field
Sunday 18 March 12
Hubble’s law
        • Expansion of the
               universe produces a
               Doppler-shift in light
               of galaxies towards
               red end of spectrum

        • The ‘redshift’ z=(λ-
               λ0)/λ0 is related to
               recessional velocity
               of each galaxy V~cz
        • V=H        0d
Sunday 18 March 12
Galaxies cluster together




Sunday 18 March 12
Comoving distance
  DC - distance between two galaxies




Density parameters

                       matter

                     dark energy
                      curvature
Sunday 18 March 12
Millenium Simulation (Springel et al. 2005)
Sunday 18 March 12
We need more redshifts

        • Measuring fundamental cosmological
               parameters depends on statistical analysis
               of large scale structure
        • A few thousand galaxies is not enough
        • Need hundreds of thousands or millions
        • Cannot do this one object at a time...
Sunday 18 March 12
Multi-Object Spectroscopy




    • Developed in late 80s/early 90s
    • Highly successful but very complex (more focus on
           getting instrument working, rather than optimising it)
Sunday 18 March 12
2dF: Two-degree Field facility
                         4-m Anglo-Australian
                              Telescope




   Lewis et al. (2002)
Sunday 18 March 12
Sunday 18 March 12
Sunday 18 March 12
Sunday 18 March 12
wavelength




Sunday 18 March 12
wavelength




Sunday 18 March 12
2dFGRS (Colless et al. 2001)




Sunday 18 March 12
N(z)~250,000!




Sunday 18 March 12
wigglez.swin.edu.au
                                              Wigglez
                     Drinkwater et al. 2010




Sunday 18 March 12
wigglez.swin.edu.au
                                                 Wigglez
                     Drinkwater et al. 2010




                                   Blake et al. 2010




Sunday 18 March 12
wigglez.swin.edu.au
                                                 Wigglez
                     Drinkwater et al. 2010




                                   Blake et al. 2010




Sunday 18 March 12
A challenging optimisation problem

 • 400 fibres to match up to N
        targets (up to ~1000)
 • Targets have priorities 1(lowest)
        to 9(highest)
 • Limited fibre reach
 • Fibres and buttons cannot
        collide, but fibre crossover ok
 •      Uniformly sample targets    [no structure imprint]

 • Prefer straighter fibres          [quicker config times]
Sunday 18 March 12
Fibre and target reach




Sunday 18 March 12
Fibre and target reach




Sunday 18 March 12
Sunday 18 March 12
Simulated Annealing
      • Donnelly et al. (1992) first proposed and
             implemented SA for field configuration, but not
             fast enough back then
      • SA simulates slow cooling of physical systems
             (e.g. glass), making small random changes at each
             temperature level
      • Metropolis (1953) algorithm determines whether
             a change is accepted
      • Fewer and fewer “bad” changes are accepted at
             lower temperatures
Sunday 18 March 12
Travelling Salesman Problem
                     Numerical Recipes (Ch. 10)




           (b) large river penalty   (c) negative river penalty!




Sunday 18 March 12
Annealing schedule
     • Start with unallocated fibres, a few hundred targets
            and an initial temperature Ti

     • Slowly cool Ti by multiplication with (1-ΔT)
     • Randomly choose new targets for each fibre,
            multiple times (up to 105 swaps per ΔT)
     • The randomisation of each fibre occurs in four ways
     • Metropolis (1953) algorithm accepts or denies each
            change, depending on global ‘quality’ of field
     • Reach quasi-static equilibrium at each temperature
Sunday 18 March 12
Four randomisation cases
before



   after




Sunday 18 March 12
Metropolis algorithm




Sunday 18 March 12
Metropolis algorithm


                             Boltzmann distribution in
                               statistical mechanics




Sunday 18 March 12
Objective function
                      target    close   straighten
                     priority   pairs     fibres      maximise
                                                       me!




Sunday 18 March 12
Objective function




                           Temperature




Sunday 18 March 12
A sample run


                     E




                             Temperature

Sunday 18 March 12
Simulations
  • Both uniform and clustered fields
  • Also use actual cosmological
         simulations (mock catalogues)
  • Different priority distributions
  • Fields with close pairs
  • LOTS of trial and error in
         selecting best algorithm
         parameters
  • Usually configure 1000 fields each
Sunday 18 March 12
Total target yield




Sunday 18 March 12
Total target yield




Sunday 18 March 12
Target priorities




                     lowest               highest
Sunday 18 March 12
Target priorities




                     lowest               highest
Sunday 18 March 12
Target priorities




Sunday 18 March 12
Target priorities




Sunday 18 March 12
Target priorities




Sunday 18 March 12
Uniformity
Oxford




      SA




Sunday 18 March 12
OLD
             (Oxford)




Sunday 18 March 12
NEW
             (Annealing)




Sunday 18 March 12
Fibre straightness




                     γ=0.0     γ=0.125    γ=2.0




Sunday 18 March 12
Fibre straightness




                     γ=0.0     γ=0.125    γ=2.0




Sunday 18 March 12
Algorithm summary
     • Power is in contained in the objective function
     • Performance far exceeds previous algorithms
     • Both in raw target yield and flexibility
     • Routinely used by astronomers at AAT since 2006
     • Routinely used by several large galaxy redshift surveys
     • Generic algorithm suitable to many other MOS
            instruments
     • Opportune time to apply it to MOS masks at SALT!
Sunday 18 March 12
SALT
  • Biggest single telescope in Southern Hemisphere!
  • 11.1m x 9.8m optical mirror
  • Refurbished instrumentation: April 2011
  • Second science semester starts in May 2012
  • Multi-object capability: instead of fibres, use slit-masks
  • MOS is currently being tested/commissioned
  • Perfect time to explore optimisation of mask design
photo: Lisa Crause
Sunday 18 March 12
Sunday 18 March 12
Sunday 18 March 12
MOS masks

      • Cheaper than developing a robot + fibre system
      • Use laser to cut slits in carbon fibre mask
      • Mask is placed in focal plane of telescope
      • Each slit produces a spectrum
      • Challenge is to ‘pack in’ the best arrangement of
             slits in one mask
      • A unique set of constraints c.f. fibre optimisation
Sunday 18 March 12
Laser mask cutter   MOS @ SALT
                            Slit mask cutter software GUI




Sunday 18 March 12
courtesy
                                           David
                                          Gilbank
 ~1/2 degree




                     IMACS on Magellan
                       6.5-m telescope
                            Chile
Sunday 18 March 12
courtesy
                                           David
                                          Gilbank
 ~1/2 degree




                     IMACS on Magellan
                       6.5-m telescope
                            Chile
Sunday 18 March 12
courtesy
           David
          Gilbank


                     slits




Sunday 18 March 12
courtesy
                 David
                Gilbank




Sunday 18 March 12
AIMS project
   • An exploratory study for a new mask design algorithm
   • Dr Brent Miszalski (SAAO/SALT)
   • Dr David Gilbank (SAAO)
   • Prof Bruce Bassett (AIMS/SAAO/UCT)
   • Design clear guidelines necessary for algorithm
          development to start
   • Identify most efficient and clever ways to conduct basic
          operations needed in a mask algorithm
Sunday 18 March 12
MOS mask design issues
      • What data structures to use in algorithm?
       • Hashes, vectors, lists, etc. Best choices == faster
      • How to tilt slits to capture > 1 target in field?
      • What randomisation steps to choose?
       • Shifting slit centres, extending slit size??
       • Shuffling groups of slits? Adding new slits?
      • How do we best define a “good” mask design?
       • Quantify completeness? Ensemble designs?
Sunday 18 March 12
MOS mask design issues
      • What is the best way to explore the parameter
             space of the problem?
      • Monte carlo simulations, statistics on real input data
      • Review previous MOS algorithms (especially mask
             design algorithms)
      • Most algorithms in the literature could be
             considerably improved
      • Your work could be used routinely at SALT!
Sunday 18 March 12
Applications

      • An improved MOS algorithm has multiple
             applications
      • Not just cosmological surveys (most of which are
             done on smaller telescopes with larger fields)
      • Globular clusters - spectroscopy of individual stars
      • Galaxy clusters - studying cluster properties as a
             function of redshift to bring new insights into galaxy
             formation and evolution, cosmology.

Sunday 18 March 12
Omega Centauri (ESO)




Sunday 18 March 12
Über cluster (D. Gilbank)
                              z~0.7




Sunday 18 March 12
Thank you!




        brent@saao.ac.za
Sunday 18 March 12
Sunday 18 March 12

Más contenido relacionado

Más de CosmoAIMS Bassett

Tuning your radio to the cosmic dawn
Tuning your radio to the cosmic dawnTuning your radio to the cosmic dawn
Tuning your radio to the cosmic dawnCosmoAIMS Bassett
 
A short introduction to massive gravity... or ... Can one give a mass to the ...
A short introduction to massive gravity... or ... Can one give a mass to the ...A short introduction to massive gravity... or ... Can one give a mass to the ...
A short introduction to massive gravity... or ... Can one give a mass to the ...CosmoAIMS Bassett
 
Decomposing Profiles of SDSS Galaxies
Decomposing Profiles of SDSS GalaxiesDecomposing Profiles of SDSS Galaxies
Decomposing Profiles of SDSS GalaxiesCosmoAIMS Bassett
 
Cluster abundances and clustering Can theory step up to precision cosmology?
Cluster abundances and clustering Can theory step up to precision cosmology?Cluster abundances and clustering Can theory step up to precision cosmology?
Cluster abundances and clustering Can theory step up to precision cosmology?CosmoAIMS Bassett
 
An Overview of Gravitational Lensing
An Overview of Gravitational LensingAn Overview of Gravitational Lensing
An Overview of Gravitational LensingCosmoAIMS Bassett
 
Testing cosmology with galaxy clusters, the CMB and galaxy clustering
Testing cosmology with galaxy clusters, the CMB and galaxy clusteringTesting cosmology with galaxy clusters, the CMB and galaxy clustering
Testing cosmology with galaxy clusters, the CMB and galaxy clusteringCosmoAIMS Bassett
 
Galaxy Formation: An Overview
Galaxy Formation: An OverviewGalaxy Formation: An Overview
Galaxy Formation: An OverviewCosmoAIMS Bassett
 
Spit, Duct Tape, Baling Wire & Oral Tradition: Dealing With Radio Data
Spit, Duct Tape, Baling Wire & Oral Tradition: Dealing With Radio DataSpit, Duct Tape, Baling Wire & Oral Tradition: Dealing With Radio Data
Spit, Duct Tape, Baling Wire & Oral Tradition: Dealing With Radio DataCosmoAIMS Bassett
 
From Darkness, Light: Computing Cosmological Reionization
From Darkness, Light: Computing Cosmological ReionizationFrom Darkness, Light: Computing Cosmological Reionization
From Darkness, Light: Computing Cosmological ReionizationCosmoAIMS Bassett
 
WHAT CAN WE DEDUCE FROM STUDIES OF NEARBY GALAXY POPULATIONS?
WHAT CAN WE DEDUCE FROM STUDIES OF NEARBY GALAXY POPULATIONS?WHAT CAN WE DEDUCE FROM STUDIES OF NEARBY GALAXY POPULATIONS?
WHAT CAN WE DEDUCE FROM STUDIES OF NEARBY GALAXY POPULATIONS?CosmoAIMS Bassett
 
Binary pulsars as tools to study gravity
Binary pulsars as tools to study gravityBinary pulsars as tools to study gravity
Binary pulsars as tools to study gravityCosmoAIMS Bassett
 
Cross Matching EUCLID and SKA using the Likelihood Ratio
Cross Matching EUCLID and SKA using the Likelihood RatioCross Matching EUCLID and SKA using the Likelihood Ratio
Cross Matching EUCLID and SKA using the Likelihood RatioCosmoAIMS Bassett
 
Machine Learning Challenges in Astronomy
Machine Learning Challenges in AstronomyMachine Learning Challenges in Astronomy
Machine Learning Challenges in AstronomyCosmoAIMS Bassett
 
Cosmological Results from Planck
Cosmological Results from PlanckCosmological Results from Planck
Cosmological Results from PlanckCosmoAIMS Bassett
 
D-Branes and The Disformal Dark Sector - Danielle Wills and Tomi Koivisto
D-Branes and The Disformal Dark Sector - Danielle Wills and Tomi KoivistoD-Branes and The Disformal Dark Sector - Danielle Wills and Tomi Koivisto
D-Branes and The Disformal Dark Sector - Danielle Wills and Tomi KoivistoCosmoAIMS Bassett
 
Cosmo-not: a brief look at methods of analysis in functional MRI and in diffu...
Cosmo-not: a brief look at methods of analysis in functional MRI and in diffu...Cosmo-not: a brief look at methods of analysis in functional MRI and in diffu...
Cosmo-not: a brief look at methods of analysis in functional MRI and in diffu...CosmoAIMS Bassett
 
Effective Field Theory of Multifield Inflation
Effective Field Theory of Multifield InflationEffective Field Theory of Multifield Inflation
Effective Field Theory of Multifield InflationCosmoAIMS Bassett
 

Más de CosmoAIMS Bassett (20)

Tuning your radio to the cosmic dawn
Tuning your radio to the cosmic dawnTuning your radio to the cosmic dawn
Tuning your radio to the cosmic dawn
 
A short introduction to massive gravity... or ... Can one give a mass to the ...
A short introduction to massive gravity... or ... Can one give a mass to the ...A short introduction to massive gravity... or ... Can one give a mass to the ...
A short introduction to massive gravity... or ... Can one give a mass to the ...
 
Decomposing Profiles of SDSS Galaxies
Decomposing Profiles of SDSS GalaxiesDecomposing Profiles of SDSS Galaxies
Decomposing Profiles of SDSS Galaxies
 
Cluster abundances and clustering Can theory step up to precision cosmology?
Cluster abundances and clustering Can theory step up to precision cosmology?Cluster abundances and clustering Can theory step up to precision cosmology?
Cluster abundances and clustering Can theory step up to precision cosmology?
 
An Overview of Gravitational Lensing
An Overview of Gravitational LensingAn Overview of Gravitational Lensing
An Overview of Gravitational Lensing
 
Testing cosmology with galaxy clusters, the CMB and galaxy clustering
Testing cosmology with galaxy clusters, the CMB and galaxy clusteringTesting cosmology with galaxy clusters, the CMB and galaxy clustering
Testing cosmology with galaxy clusters, the CMB and galaxy clustering
 
Galaxy Formation: An Overview
Galaxy Formation: An OverviewGalaxy Formation: An Overview
Galaxy Formation: An Overview
 
Spit, Duct Tape, Baling Wire & Oral Tradition: Dealing With Radio Data
Spit, Duct Tape, Baling Wire & Oral Tradition: Dealing With Radio DataSpit, Duct Tape, Baling Wire & Oral Tradition: Dealing With Radio Data
Spit, Duct Tape, Baling Wire & Oral Tradition: Dealing With Radio Data
 
MeerKAT: an overview
MeerKAT: an overviewMeerKAT: an overview
MeerKAT: an overview
 
Casa cookbook for KAT 7
Casa cookbook for KAT 7Casa cookbook for KAT 7
Casa cookbook for KAT 7
 
From Darkness, Light: Computing Cosmological Reionization
From Darkness, Light: Computing Cosmological ReionizationFrom Darkness, Light: Computing Cosmological Reionization
From Darkness, Light: Computing Cosmological Reionization
 
WHAT CAN WE DEDUCE FROM STUDIES OF NEARBY GALAXY POPULATIONS?
WHAT CAN WE DEDUCE FROM STUDIES OF NEARBY GALAXY POPULATIONS?WHAT CAN WE DEDUCE FROM STUDIES OF NEARBY GALAXY POPULATIONS?
WHAT CAN WE DEDUCE FROM STUDIES OF NEARBY GALAXY POPULATIONS?
 
Binary pulsars as tools to study gravity
Binary pulsars as tools to study gravityBinary pulsars as tools to study gravity
Binary pulsars as tools to study gravity
 
Cross Matching EUCLID and SKA using the Likelihood Ratio
Cross Matching EUCLID and SKA using the Likelihood RatioCross Matching EUCLID and SKA using the Likelihood Ratio
Cross Matching EUCLID and SKA using the Likelihood Ratio
 
Machine Learning Challenges in Astronomy
Machine Learning Challenges in AstronomyMachine Learning Challenges in Astronomy
Machine Learning Challenges in Astronomy
 
Cosmological Results from Planck
Cosmological Results from PlanckCosmological Results from Planck
Cosmological Results from Planck
 
Where will Einstein fail?
Where will Einstein fail? Where will Einstein fail?
Where will Einstein fail?
 
D-Branes and The Disformal Dark Sector - Danielle Wills and Tomi Koivisto
D-Branes and The Disformal Dark Sector - Danielle Wills and Tomi KoivistoD-Branes and The Disformal Dark Sector - Danielle Wills and Tomi Koivisto
D-Branes and The Disformal Dark Sector - Danielle Wills and Tomi Koivisto
 
Cosmo-not: a brief look at methods of analysis in functional MRI and in diffu...
Cosmo-not: a brief look at methods of analysis in functional MRI and in diffu...Cosmo-not: a brief look at methods of analysis in functional MRI and in diffu...
Cosmo-not: a brief look at methods of analysis in functional MRI and in diffu...
 
Effective Field Theory of Multifield Inflation
Effective Field Theory of Multifield InflationEffective Field Theory of Multifield Inflation
Effective Field Theory of Multifield Inflation
 

Último

Working together SRE & Platform Engineering
Working together SRE & Platform EngineeringWorking together SRE & Platform Engineering
Working together SRE & Platform EngineeringMarcus Vechiato
 
Structuring Teams and Portfolios for Success
Structuring Teams and Portfolios for SuccessStructuring Teams and Portfolios for Success
Structuring Teams and Portfolios for SuccessUXDXConf
 
TEST BANK For, Information Technology Project Management 9th Edition Kathy Sc...
TEST BANK For, Information Technology Project Management 9th Edition Kathy Sc...TEST BANK For, Information Technology Project Management 9th Edition Kathy Sc...
TEST BANK For, Information Technology Project Management 9th Edition Kathy Sc...marcuskenyatta275
 
Google I/O Extended 2024 Warsaw
Google I/O Extended 2024 WarsawGoogle I/O Extended 2024 Warsaw
Google I/O Extended 2024 WarsawGDSC PJATK
 
AI mind or machine power point presentation
AI mind or machine power point presentationAI mind or machine power point presentation
AI mind or machine power point presentationyogeshlabana357357
 
Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...
Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...
Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...ScyllaDB
 
Long journey of Ruby Standard library at RubyKaigi 2024
Long journey of Ruby Standard library at RubyKaigi 2024Long journey of Ruby Standard library at RubyKaigi 2024
Long journey of Ruby Standard library at RubyKaigi 2024Hiroshi SHIBATA
 
Event-Driven Architecture Masterclass: Challenges in Stream Processing
Event-Driven Architecture Masterclass: Challenges in Stream ProcessingEvent-Driven Architecture Masterclass: Challenges in Stream Processing
Event-Driven Architecture Masterclass: Challenges in Stream ProcessingScyllaDB
 
The Metaverse: Are We There Yet?
The  Metaverse:    Are   We  There  Yet?The  Metaverse:    Are   We  There  Yet?
The Metaverse: Are We There Yet?Mark Billinghurst
 
How we scaled to 80K users by doing nothing!.pdf
How we scaled to 80K users by doing nothing!.pdfHow we scaled to 80K users by doing nothing!.pdf
How we scaled to 80K users by doing nothing!.pdfSrushith Repakula
 
How Red Hat Uses FDO in Device Lifecycle _ Costin and Vitaliy at Red Hat.pdf
How Red Hat Uses FDO in Device Lifecycle _ Costin and Vitaliy at Red Hat.pdfHow Red Hat Uses FDO in Device Lifecycle _ Costin and Vitaliy at Red Hat.pdf
How Red Hat Uses FDO in Device Lifecycle _ Costin and Vitaliy at Red Hat.pdfFIDO Alliance
 
Intro in Product Management - Коротко про професію продакт менеджера
Intro in Product Management - Коротко про професію продакт менеджераIntro in Product Management - Коротко про професію продакт менеджера
Intro in Product Management - Коротко про професію продакт менеджераMark Opanasiuk
 
Oauth 2.0 Introduction and Flows with MuleSoft
Oauth 2.0 Introduction and Flows with MuleSoftOauth 2.0 Introduction and Flows with MuleSoft
Oauth 2.0 Introduction and Flows with MuleSoftshyamraj55
 
Extensible Python: Robustness through Addition - PyCon 2024
Extensible Python: Robustness through Addition - PyCon 2024Extensible Python: Robustness through Addition - PyCon 2024
Extensible Python: Robustness through Addition - PyCon 2024Patrick Viafore
 
Hyatt driving innovation and exceptional customer experiences with FIDO passw...
Hyatt driving innovation and exceptional customer experiences with FIDO passw...Hyatt driving innovation and exceptional customer experiences with FIDO passw...
Hyatt driving innovation and exceptional customer experiences with FIDO passw...FIDO Alliance
 
WebAssembly is Key to Better LLM Performance
WebAssembly is Key to Better LLM PerformanceWebAssembly is Key to Better LLM Performance
WebAssembly is Key to Better LLM PerformanceSamy Fodil
 
Continuing Bonds Through AI: A Hermeneutic Reflection on Thanabots
Continuing Bonds Through AI: A Hermeneutic Reflection on ThanabotsContinuing Bonds Through AI: A Hermeneutic Reflection on Thanabots
Continuing Bonds Through AI: A Hermeneutic Reflection on ThanabotsLeah Henrickson
 
TopCryptoSupers 12thReport OrionX May2024
TopCryptoSupers 12thReport OrionX May2024TopCryptoSupers 12thReport OrionX May2024
TopCryptoSupers 12thReport OrionX May2024Stephen Perrenod
 
Human Expert Website Manual WCAG 2.0 2.1 2.2 Audit - Digital Accessibility Au...
Human Expert Website Manual WCAG 2.0 2.1 2.2 Audit - Digital Accessibility Au...Human Expert Website Manual WCAG 2.0 2.1 2.2 Audit - Digital Accessibility Au...
Human Expert Website Manual WCAG 2.0 2.1 2.2 Audit - Digital Accessibility Au...Skynet Technologies
 
FDO for Camera, Sensor and Networking Device – Commercial Solutions from VinC...
FDO for Camera, Sensor and Networking Device – Commercial Solutions from VinC...FDO for Camera, Sensor and Networking Device – Commercial Solutions from VinC...
FDO for Camera, Sensor and Networking Device – Commercial Solutions from VinC...FIDO Alliance
 

Último (20)

Working together SRE & Platform Engineering
Working together SRE & Platform EngineeringWorking together SRE & Platform Engineering
Working together SRE & Platform Engineering
 
Structuring Teams and Portfolios for Success
Structuring Teams and Portfolios for SuccessStructuring Teams and Portfolios for Success
Structuring Teams and Portfolios for Success
 
TEST BANK For, Information Technology Project Management 9th Edition Kathy Sc...
TEST BANK For, Information Technology Project Management 9th Edition Kathy Sc...TEST BANK For, Information Technology Project Management 9th Edition Kathy Sc...
TEST BANK For, Information Technology Project Management 9th Edition Kathy Sc...
 
Google I/O Extended 2024 Warsaw
Google I/O Extended 2024 WarsawGoogle I/O Extended 2024 Warsaw
Google I/O Extended 2024 Warsaw
 
AI mind or machine power point presentation
AI mind or machine power point presentationAI mind or machine power point presentation
AI mind or machine power point presentation
 
Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...
Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...
Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...
 
Long journey of Ruby Standard library at RubyKaigi 2024
Long journey of Ruby Standard library at RubyKaigi 2024Long journey of Ruby Standard library at RubyKaigi 2024
Long journey of Ruby Standard library at RubyKaigi 2024
 
Event-Driven Architecture Masterclass: Challenges in Stream Processing
Event-Driven Architecture Masterclass: Challenges in Stream ProcessingEvent-Driven Architecture Masterclass: Challenges in Stream Processing
Event-Driven Architecture Masterclass: Challenges in Stream Processing
 
The Metaverse: Are We There Yet?
The  Metaverse:    Are   We  There  Yet?The  Metaverse:    Are   We  There  Yet?
The Metaverse: Are We There Yet?
 
How we scaled to 80K users by doing nothing!.pdf
How we scaled to 80K users by doing nothing!.pdfHow we scaled to 80K users by doing nothing!.pdf
How we scaled to 80K users by doing nothing!.pdf
 
How Red Hat Uses FDO in Device Lifecycle _ Costin and Vitaliy at Red Hat.pdf
How Red Hat Uses FDO in Device Lifecycle _ Costin and Vitaliy at Red Hat.pdfHow Red Hat Uses FDO in Device Lifecycle _ Costin and Vitaliy at Red Hat.pdf
How Red Hat Uses FDO in Device Lifecycle _ Costin and Vitaliy at Red Hat.pdf
 
Intro in Product Management - Коротко про професію продакт менеджера
Intro in Product Management - Коротко про професію продакт менеджераIntro in Product Management - Коротко про професію продакт менеджера
Intro in Product Management - Коротко про професію продакт менеджера
 
Oauth 2.0 Introduction and Flows with MuleSoft
Oauth 2.0 Introduction and Flows with MuleSoftOauth 2.0 Introduction and Flows with MuleSoft
Oauth 2.0 Introduction and Flows with MuleSoft
 
Extensible Python: Robustness through Addition - PyCon 2024
Extensible Python: Robustness through Addition - PyCon 2024Extensible Python: Robustness through Addition - PyCon 2024
Extensible Python: Robustness through Addition - PyCon 2024
 
Hyatt driving innovation and exceptional customer experiences with FIDO passw...
Hyatt driving innovation and exceptional customer experiences with FIDO passw...Hyatt driving innovation and exceptional customer experiences with FIDO passw...
Hyatt driving innovation and exceptional customer experiences with FIDO passw...
 
WebAssembly is Key to Better LLM Performance
WebAssembly is Key to Better LLM PerformanceWebAssembly is Key to Better LLM Performance
WebAssembly is Key to Better LLM Performance
 
Continuing Bonds Through AI: A Hermeneutic Reflection on Thanabots
Continuing Bonds Through AI: A Hermeneutic Reflection on ThanabotsContinuing Bonds Through AI: A Hermeneutic Reflection on Thanabots
Continuing Bonds Through AI: A Hermeneutic Reflection on Thanabots
 
TopCryptoSupers 12thReport OrionX May2024
TopCryptoSupers 12thReport OrionX May2024TopCryptoSupers 12thReport OrionX May2024
TopCryptoSupers 12thReport OrionX May2024
 
Human Expert Website Manual WCAG 2.0 2.1 2.2 Audit - Digital Accessibility Au...
Human Expert Website Manual WCAG 2.0 2.1 2.2 Audit - Digital Accessibility Au...Human Expert Website Manual WCAG 2.0 2.1 2.2 Audit - Digital Accessibility Au...
Human Expert Website Manual WCAG 2.0 2.1 2.2 Audit - Digital Accessibility Au...
 
FDO for Camera, Sensor and Networking Device – Commercial Solutions from VinC...
FDO for Camera, Sensor and Networking Device – Commercial Solutions from VinC...FDO for Camera, Sensor and Networking Device – Commercial Solutions from VinC...
FDO for Camera, Sensor and Networking Device – Commercial Solutions from VinC...
 

Optimization of Multi-Object Spectroscopy in Astronomy

  • 1. Optimisation of Multi-Object Spectroscopy in Astronomy Brent Miszalski SALT Research Fellow brent@saao.ac.za Sunday 18 March 12
  • 2. Overview • Galaxy redshift surveys • Multi-object spectroscopy (MOS) • MOS field configuration by simulated annealing • MOS at the Southern African Large Telescope (SALT) Miszalski et al. 2006, MNRAS, 371,1537 Sunday 18 March 12
  • 4. M 101 Sunday 18 March 12
  • 5. Hubble Ultra Deep Field Sunday 18 March 12
  • 6. Hubble’s law • Expansion of the universe produces a Doppler-shift in light of galaxies towards red end of spectrum • The ‘redshift’ z=(λ- λ0)/λ0 is related to recessional velocity of each galaxy V~cz • V=H 0d Sunday 18 March 12
  • 8. Comoving distance DC - distance between two galaxies Density parameters matter dark energy curvature Sunday 18 March 12
  • 9. Millenium Simulation (Springel et al. 2005) Sunday 18 March 12
  • 10. We need more redshifts • Measuring fundamental cosmological parameters depends on statistical analysis of large scale structure • A few thousand galaxies is not enough • Need hundreds of thousands or millions • Cannot do this one object at a time... Sunday 18 March 12
  • 11. Multi-Object Spectroscopy • Developed in late 80s/early 90s • Highly successful but very complex (more focus on getting instrument working, rather than optimising it) Sunday 18 March 12
  • 12. 2dF: Two-degree Field facility 4-m Anglo-Australian Telescope Lewis et al. (2002) Sunday 18 March 12
  • 18. 2dFGRS (Colless et al. 2001) Sunday 18 March 12
  • 20. wigglez.swin.edu.au Wigglez Drinkwater et al. 2010 Sunday 18 March 12
  • 21. wigglez.swin.edu.au Wigglez Drinkwater et al. 2010 Blake et al. 2010 Sunday 18 March 12
  • 22. wigglez.swin.edu.au Wigglez Drinkwater et al. 2010 Blake et al. 2010 Sunday 18 March 12
  • 23. A challenging optimisation problem • 400 fibres to match up to N targets (up to ~1000) • Targets have priorities 1(lowest) to 9(highest) • Limited fibre reach • Fibres and buttons cannot collide, but fibre crossover ok • Uniformly sample targets [no structure imprint] • Prefer straighter fibres [quicker config times] Sunday 18 March 12
  • 24. Fibre and target reach Sunday 18 March 12
  • 25. Fibre and target reach Sunday 18 March 12
  • 27. Simulated Annealing • Donnelly et al. (1992) first proposed and implemented SA for field configuration, but not fast enough back then • SA simulates slow cooling of physical systems (e.g. glass), making small random changes at each temperature level • Metropolis (1953) algorithm determines whether a change is accepted • Fewer and fewer “bad” changes are accepted at lower temperatures Sunday 18 March 12
  • 28. Travelling Salesman Problem Numerical Recipes (Ch. 10) (b) large river penalty (c) negative river penalty! Sunday 18 March 12
  • 29. Annealing schedule • Start with unallocated fibres, a few hundred targets and an initial temperature Ti • Slowly cool Ti by multiplication with (1-ΔT) • Randomly choose new targets for each fibre, multiple times (up to 105 swaps per ΔT) • The randomisation of each fibre occurs in four ways • Metropolis (1953) algorithm accepts or denies each change, depending on global ‘quality’ of field • Reach quasi-static equilibrium at each temperature Sunday 18 March 12
  • 30. Four randomisation cases before after Sunday 18 March 12
  • 32. Metropolis algorithm Boltzmann distribution in statistical mechanics Sunday 18 March 12
  • 33. Objective function target close straighten priority pairs fibres maximise me! Sunday 18 March 12
  • 34. Objective function Temperature Sunday 18 March 12
  • 35. A sample run E Temperature Sunday 18 March 12
  • 36. Simulations • Both uniform and clustered fields • Also use actual cosmological simulations (mock catalogues) • Different priority distributions • Fields with close pairs • LOTS of trial and error in selecting best algorithm parameters • Usually configure 1000 fields each Sunday 18 March 12
  • 39. Target priorities lowest highest Sunday 18 March 12
  • 40. Target priorities lowest highest Sunday 18 March 12
  • 44. Uniformity Oxford SA Sunday 18 March 12
  • 45. OLD (Oxford) Sunday 18 March 12
  • 46. NEW (Annealing) Sunday 18 March 12
  • 47. Fibre straightness γ=0.0 γ=0.125 γ=2.0 Sunday 18 March 12
  • 48. Fibre straightness γ=0.0 γ=0.125 γ=2.0 Sunday 18 March 12
  • 49. Algorithm summary • Power is in contained in the objective function • Performance far exceeds previous algorithms • Both in raw target yield and flexibility • Routinely used by astronomers at AAT since 2006 • Routinely used by several large galaxy redshift surveys • Generic algorithm suitable to many other MOS instruments • Opportune time to apply it to MOS masks at SALT! Sunday 18 March 12
  • 50. SALT • Biggest single telescope in Southern Hemisphere! • 11.1m x 9.8m optical mirror • Refurbished instrumentation: April 2011 • Second science semester starts in May 2012 • Multi-object capability: instead of fibres, use slit-masks • MOS is currently being tested/commissioned • Perfect time to explore optimisation of mask design photo: Lisa Crause Sunday 18 March 12
  • 53. MOS masks • Cheaper than developing a robot + fibre system • Use laser to cut slits in carbon fibre mask • Mask is placed in focal plane of telescope • Each slit produces a spectrum • Challenge is to ‘pack in’ the best arrangement of slits in one mask • A unique set of constraints c.f. fibre optimisation Sunday 18 March 12
  • 54. Laser mask cutter MOS @ SALT Slit mask cutter software GUI Sunday 18 March 12
  • 55. courtesy David Gilbank ~1/2 degree IMACS on Magellan 6.5-m telescope Chile Sunday 18 March 12
  • 56. courtesy David Gilbank ~1/2 degree IMACS on Magellan 6.5-m telescope Chile Sunday 18 March 12
  • 57. courtesy David Gilbank slits Sunday 18 March 12
  • 58. courtesy David Gilbank Sunday 18 March 12
  • 59. AIMS project • An exploratory study for a new mask design algorithm • Dr Brent Miszalski (SAAO/SALT) • Dr David Gilbank (SAAO) • Prof Bruce Bassett (AIMS/SAAO/UCT) • Design clear guidelines necessary for algorithm development to start • Identify most efficient and clever ways to conduct basic operations needed in a mask algorithm Sunday 18 March 12
  • 60. MOS mask design issues • What data structures to use in algorithm? • Hashes, vectors, lists, etc. Best choices == faster • How to tilt slits to capture > 1 target in field? • What randomisation steps to choose? • Shifting slit centres, extending slit size?? • Shuffling groups of slits? Adding new slits? • How do we best define a “good” mask design? • Quantify completeness? Ensemble designs? Sunday 18 March 12
  • 61. MOS mask design issues • What is the best way to explore the parameter space of the problem? • Monte carlo simulations, statistics on real input data • Review previous MOS algorithms (especially mask design algorithms) • Most algorithms in the literature could be considerably improved • Your work could be used routinely at SALT! Sunday 18 March 12
  • 62. Applications • An improved MOS algorithm has multiple applications • Not just cosmological surveys (most of which are done on smaller telescopes with larger fields) • Globular clusters - spectroscopy of individual stars • Galaxy clusters - studying cluster properties as a function of redshift to bring new insights into galaxy formation and evolution, cosmology. Sunday 18 March 12
  • 64. Über cluster (D. Gilbank) z~0.7 Sunday 18 March 12
  • 65. Thank you! brent@saao.ac.za Sunday 18 March 12