SCIENTIFIC ACTIVITIES - Fields Institute - University of Toronto

Distinguished Lecture Series in Statistical Sciences
JIANQING FAN IS WELL-KNOWN FOR HIS WORK
in financial econometrics, computational biology, semiparametric
and nonparametric modeling, and other aspects of statistical
theory and methodologies. He is the Frederick I. Moore Class
of 1918 Professor in Finance, Director of the Committee of
Statistical Studies at Princeton, the Past President of the Institute
of Mathematical Statistics, and winner of the 2000 COPSS
Presidents’ Award.
In early May,
Jianqing gave the
Distinguished Lecture
Series in Statistical
Science at the Fields
Institute. His two
lectures, titled Vastdimensionality
and
sparsity and ISIS:
A vehicle for the
universe of sparsity,
focused primarily on
high-dimensional statistical modelling and feature selection.
These aspects became important with the advent of mass data
collection, advances in computation, and the discovery of new
interplay between various natural and social sciences. In his public
lecture, Fan outlined the problem of high dimensionality in fields
Coxeter Lecture Series
THE 2010 SUMMER THEMATIC PROGRAM ON THE
Mathematics of Drug Resistance in Infectious Diseases was held at
the Fields Institute during July and August. In association with
this thematic program, Professor Neil Ferguson was invited to
the Fields Institute to deliver the Coxeter Lecture Series on
Mathematical modelling of emerging infectious disease epidemics and their
control.
Ferguson, a Professor of Mathematical Biology in the
Division of Epidemiology, Public Health, and Primary Care of the
Medical School at Imperial College, is a world leader in the use of
mathematical models in infectious disease epidemiology. He is the
Director of the MRC Centre for Outbreak Analysis and Modelling.
In the first lecture, Ferguson reviewed the development of
outbreak modelling over the last two decades and discussed the
drivers which lead to more complex computational simulations
being increasingly used replacing simpler compartmental models of
disease transmission. The second lecture discussed ways in which
modelling can be optimally used to assist public health policymakers
in their planning for and reaction to emerging infectious disease
threats—an issue on which Ferguson is an expert, and which was
14 FIELDSNOTES | FIELDS INSTITUTE Research in Mathematical Sciences
Vast-dimensionality and Sparsity
as diverse as bioinformatics, genetics, physics, and economics,
discussing the twin challenges of noise accumulation and
spurious correlations. The notion of sparsity or, more generally,
homogeneity, was methodically described for feasible inference
to high-dimensional problems. After an overview of the penalized
likelihood approach for variable selection, Fan communicated the
idea of large-scale screening and moderate-scale selection together
with conditional inference as an effective solution to highdimensional
problems. Using this approach in an analysis of U.S.
housing price indices over a 30-year period, Fan demonstrated
impressive prediction improvements.
His second presentation was meant for a more specialized
audience. Exploiting sparsity, Fan outlined a unified framework
for solving high-dimensional variable selection problems. He
applied iterative vast-scale screening followed by moderate-scale
variable selection, resulting in a process called ISIS. Applying this
process to multiple regression, generalized linear models, survival
analysis, and machine learning, Fan demonstrated its overall
reach via marginal variable screening and penalized likelihood
methods. With tailored simulation studies and manipulation of
empirical data from disease classifications and survival analyses,
Fan demonstrated the advantages of a folded-concave over convex
penalty method on sure screening properties, false selection sizes
and model selection consistency.
Elif Fidan Acar (Toronto)
Mathematical Modelling of Emerging Infectious Diseases
of great interest to the thematic program participants. The third
lecture focused on the potential impact of antiviral resistance
during an influenza pandemic. He offered several explanations
for new findings that show the degree to which previous risk
assessments concerning antiviral resistance in influenza pandemics
have been over-pessimistic. In the lecture, Ferguson touched on
the critical issue of the dependence of the final impact of resistance
during a closed epidemic on the transmissibility of a sensitive
and resistant virus, the mutation rate from one type to the other,
and the level of seeding of both viral types at the beginning of
the epidemic. He argued that resistance is not likely to entail a
substantial reduction of effectiveness of antivirals during the start
of a pandemic, but that intensive drug use in this phase can lead
to a higher degree of resistance in later epidemics. His concluding
remark that “simple models suggest antiviral resistance could be a
major issue in the first wave of a new pandemic, but allowing for
spatial heterogeneity reduces speed of resistance” strongly echoed
the theme of transmission heterogeneity of the two-week block of
this entire thematic program on mathematics for drug resistance.
Jianhong Wu (York)