WOGAS3 Programme and Working Groups
WOGAS3 Programme
Below is a detailed programme for the WOGAS3 workshop. Preliminary material and other information for the working groups themselves follow at the end of the programme.
The meeting is taking place in Complexity, Zeeman (Maths/Stats) Building, University of Warwick, with the talks happening in the Complexity Seminar Room (D1.07).
Tuesday 5th April
am: Arrivals/coffee
12.30: Buffet lunch
1.25: Opening Remarks: Frank Critchley (Open University)
1.30: Opening session:
- 1.30: Eva Riccomagno (Genova): Algebraic Statistics . an overview (abstract to follow)
- 2.20: Paul Marriott (Waterloo): Information Geometry . an overview (abstract to follow)
- 3.10: Giovanni Pistone (Carlo Alberto): Algebraic Statistics and Information Geometry: two examples
4.00: Tea
4.30: Session 1:
- 4.30: Helene Gehrmann (Oxford): Graphical Gaussian models with symmetries and regular colourings
- 5.00: Sofia Massa (Oxford): Algebraic aspects of structural meta-analysis
- 5.30: Paul Byande (Montpellier 2): Affine structures and information geometry
- 6.00: Hiroyuki Nakahara (Riken): Hierachical interaction structures of neural coding, in relation to information geometry
- 6.45: Session Ends
6.45: Reception
7.45: Dinner at Rootes (for those staying in Warwick Accommodation)
Wednesday 6th April
9.00: Session 2:
- 9.00: Hélène Massam (York, Canada): Geometry of hierarchical discrete loglinear models for Bayes factors
- 9.45: Johannes Rauh (MPI): Maximising the information divergence
10.30: Coffee
11.00: Session 3: Working Group: Geometry of Model Space
- 11.00: Frank Critchley (Open University): Computational information geometry for statistical science . theory and practice
- 11.45: Vahed Maroufy (Waterloo): Geometry of local mixture models
- 12.05: Zhiyue Huang (Waterloo): Sparsity in the geometry of mixture models
- 12.25: Discussion, led by convenors
1.00: Lunch
2.30: Session 4: Working Group: Graphical Models with potential AS/IG synergies
- 2.30: Piotr Zwiernik (Mittag-Leffler Institute): Why do we care about inequalities in (algebraic) statistics?
- 3.00: Robin Evans (Washington): Smoothness of discrete conditional independence models
- 3.30: Discussion, led by convenors
4.30: Tea
5.00: Session 5: Working Group: Decision Geometry
- 5:00 Phil Dawid (Cambridge): A geometric approach to Bayesian local sensitivity (joint work with Monica Musio, Cagliari)
- 5.30: Discussion, led by convenors
- 6.30: Session ends
7.00: Dinner
Thursday 7th April
9.00: Session 6:
- 9.00: Shun-Ichi Amari (Riken): Conformal changes of divergences and information geometry
- 10.00: Karim Anaya-Izquierdo (Open University): A geometric treatment of overdispersion
10.30: Coffee
11.00: Session 7: Working Group: Hybrid symbolic-numerical approaches
- 11.00: Hugo Maruri Aguilar (Queen Mary): Betti numbers and design of experiments
- 11.30: Kei Kobayashi (ISM): Algebraic computations for asymptotically efficient estimators via information geometry
- 12.00: Mark Richardson (Oxford): Numerical computing with Chebfun [demo file]
- 12.30: Paul Marriott (Waterloo): (title/abstract to follow)
- 12.40: Discussion, led by convenors
1.00: Lunch
2.00: Session 8: Working Group: Geometry of MCMC methods
- 2.00: Julien Cornebise (UCL): Exploiting geometry to design MCMC methods for nonlinear dynamic systems
- 2.20: Vassilios Stathopoulos (Glasgow): MCMC sampling of finite mixture and admixture models: when will geometry help?
- 2.40: Krzysztof Latuszynski (Warwick): Robustness of MALA algorithms
- 3:00 Discussion, led by convenors
4.00: Closing remarks
4.30: Tea and departures
Working Group Information
WOGAS 3 will focus on several working group areas, which are listed below. Information relating to each of those groups will also be posted here as it becomes available.
o Decision Geometry - Organisers: Phil Dawid and Matt Parry
Paper for Discussion [PDF]
o Geometry of MCMC - Organisers: Mark Girolami and Matt Parry
Three 20 min talks on
a. Exploiting geometry to design MCMC methods for nonlinear dynamic systems - Julien Cornebise, UCL
b. MCMC sampling of finite mixture and admixture models: when will geometry help? - Vassilios Stathopoulos, Glasgow
c. Robustness of MALA algorithms - Krzysztof Latuszynski, Warwick
Some Relevant work:
Riemann Manifold Langevin and Hamiltonian Monte Carlo
JRSSB Discussion Paper by Girolami and Calderhead to appear in March [Wiley JRSSB Webpage]
o Graphical Models, with potential AS/GS synergies - Organisers: Frank Critchley and Jim Smith
Discussion document here [PDF]
Three Talks on Computational Information Geometry for discussion: [CIG1] [CIG2 - PDF,12Mb, ZIP, 2Mb] [CIG3]
o Hybrid symbolic-numerical approaches - Organiser: Henry Wynn
Discussion document here [PDF]
o Geometry of Model Space - Organiser: Paul Marriott
Discussion document here [PDF]