GAMBIT: Towards a Global And Modular Beyond-the-Standard-Model Inference Tool (15w5115)

Arriving in Oaxaca, Mexico Sunday, September 27 and departing Friday October 2, 2015


(Imperial College)

Martin White (Centre of Excellence for Particle Physics at the Terascale)


The primary goals of this BIRS 5-day Half Workshop will be:

  1. To re-evaluate the computational and statistical techniques used to carry out global fits in particle and astroparticle physics.
  2. To develop a second generation global-fitting package for particle and astroparticle physics: GAMBIT, the Global And Modular Beyond-the-Standard-Model Inference Tool.
  3. To determine the statistical approach(es) to be supported by GAMBIT.
  4. To identify new computational strategies that optimize the efficiency and accuracy of GAMBIT relative to first-generation codes.
  5. To prioritize and discuss different physics analyses to be carried out with GAMBIT.

Relevance, importance and timeliness

Existing global fits [IC22Methods, Fittino12, Mastercode12b, Roszkowski12, Strege13] cover only a very small subset of interesting particle models; most have dealt with only the very simplest versions of supersymmetry. This is partly for computational reasons, as efficiently exploring the parameter spaces of more complicated models is extremely time consuming. Existing optimization and inference techniques are barely capable of dealing with even the models that have been considered so far [Akrami09,SBSpike]. Efforts to date have rarely been truly `global', as the full range of possible observables and datasets (e.g. indirect searches for dark matter) have not been included in a detailed way. The present generation of global analysis suites will all hit a brick wall within the next year in their abilities to deal with alternative theories, additional observables and the advanced numerical and statistical algorithms required for producing genuinely robust results.

Future progress in understanding which BSM models are favoured by experimental data will be contingent upon massively expanding the range of theories to which global fits have been applied, and the number of experimental results included in them. The only way to do this is to reconsider the computational and statistical tools used to carry them out, from the ground up. Our BIRS Workshop will do exactly this.

By rewriting the computational and statistical bedrock of BSM global fits, we will be in a position to discuss and develop a new, second-generation global fitting suite for particle and astroparticle physics: GAMBIT. GAMBIT will transform the budding field of global fits, by providing a framework in which new theories, observables, likelihoods and scanning algorithms can be quickly, easily and consistently combined in order to completely and rigorously test essentially textit{any} proposed extension of particle physics beyond the Standard Model.


Creating a tool such as GAMBIT is an extremely demanding problem in applied statistics and applied computer science, requiring the use of many specialized statistical techniques and computer codes. It is also a highly interdisciplinary physics problem, straddling the theoretical and experimental branches of both astronomy and particle physics. Researchers from such leading institutes as the Oskar Klein Centre for Cosmoparticle Physics (Stockholm), the Imperial Centre for Inference and Cosmology (London) and the Centre of Excellence for Particle Physics at the Terascale (Adelaide, Melbourne, Sydney) have all already expressed their extremely strong interest in taking part in the Workshop. The combination of guests we envisage for this Workshop is unique, bringing together the broad range of expertise in astrophysics and particle physics, theory and experiment, DM phenomenology, statistical and numerical methods required to make our ambitious vision a reality.

By collecting world experts in all these areas for an intensive, focused week on the problems, strategy and first attempted implementation of a second-generation global-fitting code, the BIRS GAMBIT Workshop will provide new opportunities for exchange of ideas between members of all these fields. This will drive the field of BSM global fits forward by enabling new collaborations. The Workshop we hold at BIRS will form the basis for what we hope will be an ongoing formal consortium of experts in the BSM global fitting arena, dedicated to the development and usage of the GAMBIT global-fitting framework. If this proposal is successful, given the long lead-time for holding a Workshop at BIRS and the urgency of the problem, we will attempt to begin preliminary discussions by email before even arriving at BIRS. This will ensure that we hit the ground running in Banff, and come away from the Workshop having achieved the maximum progress possible towards our Objectives.

The GAMBIT BIRS Workshop, and the consortium that we hope will rise from it, will also prove a fertile training ground for upcoming talent in the field. We have identified a number of extremely promising doctoral students and postdoctoral researchers with an interest in BSM global fitting; their involvement in the Workshop and consortium will give them unrivalled experience in the ingredients of BSM global fits, and an excellent opportunity to influence the future directions of the field, all but ensuring that they become its future leaders.


The first two days of the Workshop will consist of presentations. Some of these will evaluate existing statistical frameworks (e.g. the composite profile likelihood and Bayesian posterior), computational methods (e.g. nested sampling) and overall code structures used in BSM global fits. We will encourage those presenters to approach their subject matter with a critical, idealistic eye, with the goal of identifying how a global-fitting framework should be designed so as to reflect best practices in these areas -- rather than just the `simplest thing to get running'. This will address Objective 1. Other presentations will give proposals for using new methods in BSM global fits, like Hamiltonian sampling, random forests and differential evolution. This will address Objectives 3 and 4. A third set of presentations will focus on specific aspects of the physics that should be addressed by future BSM global fits: dark matter, collider and flavour physics. This will address Objective 5.

The latter three days of the Workshop will be devoted almost entirely to collaborative development of the GAMBIT suite, including group coding sessions and code design mini-workshops. Towards the end of the week, we also envisage some code presentations and tutorials, where a small number of attendees show the group as a whole the progress they have made, and how others can use or expand on it. These will address Objective 2.


  • [Bertone05]
    G.~{Bertone}, D.~{Hooper}, and J.~{Silk}, {it {Particle dark matter: evidence,
    candidates and constraints}}, {em physrep} {bf 405} (2005) 279--390,
    [href{}{{tt hep-ph/0404175}}].

  • [BaerTata]
    H.~{Baer} and X.~{Tata}, {em {Weak Scale Supersymmetry}}.
    newblock Cambridge University Press, 2006.

  • [Degrassi12]
    G.~{Degrassi}, S.~{Di Vita}, {em et.~al.}, {it {Higgs mass and vacuum
    stability in the Standard Model at NNLO}}, {em jhep} {bf 8} (2012) 98,
    [href{}{{tt arXiv:1205.6497}}].

  • [CRESST11]
    G.~{Angloher}, M.~{Bauer}, {em et.~al.}, {it {Results from 730 kg days of the
    CRESST-II Dark Matter search}}, {em epjc} {bf 72} (2012) 1971,
    [href{}{{tt arXiv:1109.0702}}].

  • [Bringmann12]
    T.~{Bringmann}, X.~{Huang}, A.~{Ibarra}, S.~{Vogl}, and C.~{Weniger}, {it
    {Fermi LAT search for internal bremsstrahlung signatures from dark matter
    annihilation}}, {em jcap} {bf 7} (2012) 54,
    [href{}{{tt arXiv:1203.1312}}].

  • [Pamelapositron]
    O.~{Adriani}, G.~C. {Barbarino}, {em et.~al.}, {it {An anomalous positron
    abundance in cosmic rays with energies 1.5--100,GeV}}, {em nat} {bf 458}
    (2009) 607--609, [href{}{{tt

  • [IC22Methods]
    P.~{Scott}, C.~{Savage}, J.~{Edsj{"o}}, and {the IceCube Collaboration:
    R.~Abbasi et al.}, {it {Use of event-level neutrino telescope data in global
    fits for theories of new physics}}, {em jcap} {bf 11} (2012) 57,
    [href{}{{tt arXiv:1207.0810}}].

  • [Fittino12]
    P.~{Bechtle}, T.~{Bringmann}, {em et.~al.}, {it {Constrained supersymmetry
    after two years of LHC data: a global view with Fittino}}, {em jhep} {bf
    6} (2012) 98, [href{}{{tt

  • [Mastercode12b]
    O.~{Buchmueller}, R.~{Cavanaugh}, {em et.~al.}, {it {The CMSSM and NUHM1 in
    light of 7 TeV LHC, B $_{ s }${$to$} {$mu$} $^{+}$ {$mu$} $^{-}$ and
    XENON100 data}}, {em epjc} {bf 72} (2012) 2243,
    [href{}{{tt arXiv:1207.7315}}].

  • [Roszkowski12]
    L.~{Roszkowski}, E.~M. {Sessolo}, and Y.-L.~S. {Tsai}, {it {Bayesian
    implications of current LHC supersymmetry and dark matter detection searches
    for the constrained MSSM}}, {em prd} {bf 86} (2012) 095005,
    [href{{arXiv:1202.1503}}{{tt {arXiv:1202.1503}}}].

  • [Strege13]
    C.~{Strege}, G.~{Bertone}, {em et.~al.}, {it {Global fits of the cMSSM and
    NUHM including the LHC Higgs discovery and new XENON100 constraints}}, {em
    jcap} {bf 4} (2013) 13,
    [href{{arXiv:1212.2636}}{{tt {arXiv:1212.2636}}}].

  • [Akrami09]
    Y.~{Akrami}, P.~{Scott}, J.~Edsj{"o}, J.~{Conrad}, and L.~{Bergstr{"o}m},
    {it {A profile likelihood analysis of the Constrained MSSM with genetic
    algorithms}}, {em jhep} {bf 4} (2010) 57,
    [href{}{{tt arXiv:0910.3950}}].

  • [SBSpike]
    F.~{Feroz}, K.~{Cranmer}, M.~{Hobson}, R.~{Ruiz de Austri}, and R.~{Trotta},
    {it {Challenges of profile likelihood evaluation in multi-dimensional SUSY
    scans}}, {em jhep} {bf 6} (2011) 42,
    [href{}{{tt arXiv:1101.3296}}].