FCT

R&D Institutions

Resultado da avaliação 2007 na área de Matemática

Unidade de I&D

Centro de Matemática [MATH-Centro-Covilha-212] visitada em 22/07/2008

Classificação: Good

Comentários do painel de avaliação
Sobre a unidade
The rating should be read as a GOOD minus. This unit has suffered from the loss of senior researchers since the last evaluation. Currently it has 17 PhD researchers.

Until 2006 comprised a single group in statistics, and the group in computation had a different focus. After reorganization in 2006, the research activity is structured into five groups:

1. Algebra, topology and geometry
2. Extreme value theory
3. Models of physical processes in continua
4. Modelling, computing, and inference
5. Statistical natural language processing

The groups in extreme value theory and in models of physical processes have the strongest publication records, but mostly in specialist journals, some of low impact. Overall one senses a lack of ambition in the choice of the research topics and the outlets. The unit suffers from its physical isolation, which makes it imperative that its members travel to leading centres abroad and actively seek a wider range of national and international contacts.

The report of the 2002 external review highlighted persistent serious problems with this unit, which subsist. Some comments from 2002 might be applied again verbatim. In particular:

--Publications: the productivity of some groups remains low, and most of the papers appear in national or poorly-ranked international journals. The panel was surprised to learn that PhD theses at this unit are still written in Portuguese, which is inadequate preparation for an academic career.

--Students: the geographical isolation of the unit and the poor networking of some of its members lead to its poor overall training record, though this has improved. If the isolation is to be overcome, it is essential that future members of the unit undergo training abroad.

--Vision: The global vision and scientific strategy of the unit need revision. The number of seminars has increased from its former low base, but supplementing external speakers with a working seminar or journal club, especially in statistics, might help to forge a common outlook.

--Funding: All groups should actively pursue engagement in national networks and partnerships, which will help to overcome its isolation.

--Presentation: The presentation was greatly improved over the last mathematics panel visit, but could still benefit from comparison with those of other units, such as in Coimbra. In particular it was difficult to get a synthetic view of the unit, and of key elements such as MSc and PhD training; owing to this some of the numbers given below may be slightly incorrect.

The previous panel recommended continuation of this unit (then called a centre) only for three years from 2002, with the stipulation that if there were no major changes to address the main issues mentioned above, then the unit should be closed and its research-active members dispersed elsewhere. In the meantime some research-active scientists have arrived, and this panel is not minded to support the previous recommendation at this time. However strenuous efforts are needed to address the above difficulties if the unit is to avoid dissolution.
Sobre os grupos de investigação
AGT-Algebra, Geometry, Topology [RG-MATH-Centro-Covilha-212-2969]
1. Algebra, topology and geometry
This group has recently grown from 2 to 4 PhD researchers, and has interesting projects in differential geometry and in algebra. However it appears to lack in international recognition and some of its members lack publications in high-level peer-reviewed journals. Others publish regularly in known mathematical journals. There are currently 2 MSc and 1 PhD students, but the group appears to have had only one MSc student in the previous funding round. The panel hopes that the leadership of the group will shortly change to reflect the interests of its new members.
ATG - Algebra, Topology, Geometry [RG-X-MATH-Centro-Covilha-212-2970]
DECONT- Differential Equations of Continuum Mechanics: Solvability and Localization Effects, Numerical Methods [RG-X-MATH-Centro-Covilha-212-1306]
EVT – Extreme Value Theory [RG-MATH-Centro-Covilha-212-1310]
2. Extreme value theory
The second group of 3 PhD researchers is focused on probabilistic topics in statistics of extremes, and contains active specialists who publish in respected international journals. The research proposal is worthwhile, as extremal properties of many stochastic models used in practice remain to be worked out in detail. The group currently has 2 PhD and 1 MSc students, and over the past period its members produced 1 PhD, now a member of the group. The panel recommends that the group build international contacts, in order to broaden their range of topics and increase their visibility.
EXMIXMOD - Extremes and Mixed Models [RG-X-MATH-Centro-Covilha-212-1308]
IRTMOD-Item Response Theory Models [RG-X-MATH-Centro-Covilha-212-1312]
MPPC – Models of Physical Processes in Continua [RG-MATH-Centro-Covilha-212-1311]
3. Models of physical processes in continua
This group has 4 PhD researchers, not all of whom have published in recent times. Overall the work is of higher quality than the journals in which it appears. The objectives of the group include numerical studies, but there is no indication of work in this area. There are international contacts, but overall the research agenda seems somewhat out-dated. The past PhD training effort was respectable for a group of this size, but depended heavily on its former senior members, and continues to do so. This group will become sub-critical when the remaining senior member retires. The panel recommends its absorption into a larger unit elsewhere.
SEASON - Semantics and Structure of Languages [RG-MATH-Centro-Covilha-212-1315]
STATISTICA – Modelling, Computing, Inference [RG-MATH-Centro-Covilha-212-1314]
4. Modelling, computing, and inference
This group of 5 PhD researchers energetically undertakes statistical modelling of educational studies both in Portugal and Brazil (many of the group's publications are in Portuguese) but should provide more published evidence of useful methodological innovation. At present there are no publications in high-level journals. The group has external funding for national educational studies and may have potential for high impact in this domain, but the panel is unable to judge this. There are currently 5 MSc and 5 PhD students, but the likely quality of their work is a cause for concern, given the publication record of the group. The panel recommends that the group should build and maintain long-term collaborations with leading institutions in social and educational statistics, if it wishes to make more methodological impact within statistics.
STATNAT - Statistical Natural Language Processing [RG-MATH-Centro-Covilha-212-2966]