Caros colegas,
numa ação conjunta, o Departamento de Estatistica da USP
e da UFMG estão trazendo ao Brasil o pesquisador
Stephane Robin. Ele e' especialista na analise e modelagem
estatistica de dados de redes (web, biologicas, sociais, etc.).
Ele já deu uma conferência no IME-USP na segunda-feira e dara' tres
conferencias no departamento de estatistica da UFMG nos dias e
horarios especificados a seguir. Infelizmente, eu já deveria ter anunciado
na ABE-L os seminarios e assim lamento o inconveniente da mensagem
atrasada.
Para quem tiver interesse, a sua pagina web e' a seguinte:
http://www.agroparistech.fr/-Stephane-Robin-.html
Os seminarios são abertos a todos.
Renato Assuncao
******************************************************************************
1 – Some Statistical Aspects of Interaction Network Analysis
Dia 20 de novembro, Quinta-feira, de 11 as 12:30
Sala 2076 - ICEx
In this lecture, some random graph models will be presented in
the perspective of statistical inference for the underlying
parameters. The talk will provide a (very partial) review of
published papers, focusing on interaction networks. The
modelling of the global topology of the network will be first
considered and extended to account for available covariates.
The analysis of local characteristics, such as edge prediction
of motif occurrence will then be presented. The last part of
the presentation will address network inference in the context
of Gaussian graphical models.
******************************************************************************
2 – Uncovering Structure in Interaction Networks
Dia 21 de novembro, Sexta-feira, de 11 as 12:30
Sala 2076 - ICEx
Many observed network present a modular topology. Furthermore,
nodes often present heterogenous connection profiles. Mixture
models provide a natural framework to account for such
heterogeneity, but raise difficult inference problems because
of very intrincate dependencies due to the network structure. A
variationnal approximation will be presented, which allows to
get (appproximate) maximum likelihood estimates. Some applications
of this model to various fields (biology, sociology, ecology) will
then be presented. Possible extensions of the inference strategy
will be discussed in the end.
******************************************************************************
3 – Detecting Exceptionnal Motifs in Interaction Networks
Dia 21 de novembro, Sexta-feira, 15:30
Sala 2076 - ICEx
Local structure of a network may help to understand its global
behaviour. In many networks, some specific patterns (triangles,
loops, stars, ...) are observed very frequently. We will consider
the statistical assessment of the significance of these frequencies.
The problem will be introduced via motif detection in random
sequences to give some intuition of the strategy that will be then
developped in the network context. Some application to biological
networks will be provided. The choice of a 'relevent' network
model will also be discussed.