Hidden Markov Models and Complex Systems Programme
Under the auspices of an NZIMA (New Zealand Institute of Mathematics and its Applications) Programme, a series of workshops and seminars were conducted throughout New Zealand in the second half of 2005, bringing together statistical practitioners who use Hidden Markov Models and related methods to analyse data from a wide variety of applications. The aim of the Programme was to augment the research programmes of New Zealand scientists working with HMM methods.
INFORMAL WORKSHOP
It is proposed to hold a workshop on Hidden Markov Models at the Royal Society's rooms in Wellington on Wednesday 3 September 2008, as a 1-day extension, or satellite meeting, of the NZSA Conference in Hamilton, 1-2 September 2008, at which a HMM session is also planned. There is no charge for participants.
The aims are to provide a follow-up meeting to the earlier NZIMA-funded programme on ‘Hidden Markov Models and Complex Systems’, and to take advantage of Walter Zucchini’s visit to Wellington as the Shayle Searle Lecturer for 2008. Walter will give an invited lecture on "Latent-state models with feedback − an extension of hidden Markov models".
Click for Abstract to Walter's Talk
We want to try to organize the Workshop so that it can provide an opportunity to see what directions the work on hidden Markov models in NZ has been taking over the last few years. We shall try to keep the sessions, and indeed the whole Workshop, at an informal level, with an emphasis on the application and the character of the model structure used.
Invitation to Participate
There will be no formal registration process for the Workshop. The one vital thing we will need, however, is a list of participants, and those of you who would like or be willing to give a short presentation. Please notify your interest in participating by emailing Dr. Yuichi Hirose, and particularly your interest in presenting a short paper or contribution to discussion.Programme
We have in mind talks of 20-25 minutes or so, with enough time for some discussion, outlining an application area or any new developments in methodology or structure. We also hope to have time for a question and answer session with Walter and others joining in. If we have your suggestions in enough time, we will try to group them into themes, as suggested below or in other ways, and send out a draft programme around mid-August.Themes
Papers will be welcome on any aspect of hidden Markov models. Some themes which our own work has led us to consider, and which might serve as examples of the sorts of topics which seem to be emerging, are as follows:Ancilliary Variables: One of the features of which we have become increasingly aware in recent years is the variety of ways in which hidden Markov structure can be linked to ancillary variables. These may form a vector structure, including spatial aspects, and be included as part of a complex observation process. Or the ancillary variables may influence the transition structure of the hidden chain, or the distribution of the observed variables.
Intervals of a Point Process: Another theme which has emerged from our experiences with fitting hidden Markov models in a point process context is that there is very little loss of precision in looking at the process only as the sequence of occurrence points, and hence as a discrete-time hidden Markov process indexed by the intervals between occurrence points.
Generalisations: Another theme again concerns the extension of the hidden Markov algorithms to more general process structures such as random graphs.
Visiting Lecturer
Prof Walter Zucchini, University of Göttingen, is well-known for his work on hidden Markov models, including his co-authorship of Hidden Markov and Other Models for Discrete-valued Time Series (MacDonald, I. L. and W. Zucchini, 1997; London: Chapman & Hall). His course (Hidden Markov Models Short Course, Zucchini, W., 2005, 3–4 April 2005, Macquarie University, Sydney) formed the basis for David Harte's R package for Hidden Markov Models. Walter had worked on many fields of application, including meteorology, ecology, finance, animal behaviour, physiology and forestry. Walter will give an invited lecture on "Latent-state models with feedback − an extension of hidden Markov models".
Venue
The Workshop will be held in the large lecture room in the Royal Society’s Rooms at 11 Turnbull St, Thorndon, Wellington, from 9 am - 4 pm. There is no charge for participants. Morning and afternoon teas will be available, and there are several neighbourhood cafes where lunch can be obtained. The Workshop is jointly sponsored by the NZSA, Victoria University of Wellington and Statistics Research Associates.Contact
For further enquiries about the workshop contact Dr. Yuichi Hirose. For feedback about the webpage contact Roger Littlejohn.
