UCL, ISM, ROIS and MEXT-jointly-supported workshops, STM2014 and CSM2014 have been successfully finished. The slides and videos for some talks are available on "Technical Program" page.

Thank you very much! 

  The analysis of complex and massive data sets which display attributes of spatial and temporal characteristics is a growing field of research. Traditionally the two fields have been treated predominantly via a range of different approaches, depending on the discipline in which the applications are under study. For instance in spatial statistics and geo-statistics there is a long history of spatial modeling via regression models, random fields and parametric approaches. There is also a large literature on spatial extremes and the analysis of such features of spatial temporal data. In machine learning there are new approaches being considered based on semi-parametric and non-parametric modeling paradigms which incorporate Bayesian modeling. Then in the signal processing and engineering communities there have been a range of frequency and spatial methods developed based on filters, parametric modeling, regression - basis expansion models, wavelets regression models, and splines.
  Recently there have been new methods developed for the analysis of spatial and temporal data based on state space modeling, stochastic geometry, asymptotic approximations/expansions, quantile regressions and spatial temporal dependence structures that are constructed via copulas. The estimation of such models has significantly advanced through ideas such as INLA, LAPLACE, EP and estimation via Monte Carlo strategies such as sequential Monte Carlo and Markov chain Monte Carlo sampling.
  In this workshop we aim to introduce these approaches for a range of real applications in areas of machine learning, speech processing, wireless communications modeling, sensor networks and classification.

- Modern state space models and estimation
- Tensor and sparse modelling structures
- Long range percolation and network connectivity
- Temporal/Frequency domain modelling and application
- Bayesian non-parametric modeling for speech and language processing applications

- High dimensional spatial and temporal Bayesian inference methodology based on combination of Markov chain Monte Carlo methods and Sequential Monte Carlo methods
- Spatial and frequency based dependence modeling via copula frameworks
- Spatial and temporal heavy tailed processes with particular focus on working with multivariate and spatial stable processes

Conference Room 1 (2F), Institute of Statistical Mathematics, Tokyo, Japan

28/July/2014 (Mon) - 29/July/2014 (Tue): STM2014
30/July/2014 (Wed) - 31/July/2014 (Thu): CSM2014

Invited speakers
- Prof. Nourddine Azzaoui (Université Blaise Pascal)
- Prof. Jen-Tzung Chien (National Chiao Tung U.)
- Prof. Arnaud Doucet (Oxford U.)
- Prof. Norikazu Ikoma (KIT)
- Prof. Kenji Fukumizu (ISM)
- Prof. Konstatin Markov (Aizu U.)
- Prof. Tomoko Matsui (ISM)
- Prof. Daichi Mochihashi (ISM)
- Prof. Pierre Del Moral (UNSW)
- Prof. Tor Andre Myrvoll (SINTEF)
- Dr. Ido Nevat (Institute for Infocom Research, A-Star)
- Prof. Yosihiko Ogata (ERI, U. Tokyo and ISM, ROIS)
- Prof. Daniel P. Palomar (HKUST)
- Dr. Gareth Peters (UCL)
- Prof. Francois Septier (Telecom lille 1)
- Prof. Taiji Suzuki (Tokyo Institute of Technology)
- Prof. Kazuya Takeda (Nagoya U.)
- Prof. Mario Wüthrich (ETH Zurich)

The working language of the conference is English.


- Research Organization of Information and Systems (ROIS)
- Research Center for Statistical Machine Learning, Institute of Statistical Mathematics (ISM)
- UK Royal Society International Exchange Grant
- Research Organization of Information and Systems (ROIS)
- MEXT undertake project "Cooperation with Math Program"
- UK Royal Society International Exchange Grant