BEGIN:VCALENDAR
PRODID:-//PSU Mathematics Department//Seminar iCalendar Generator//EN
VERSION:2.0
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:Probability and Financial Mathematics Seminar
X-WR-TIMEZONE:America/New_York
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20150828T153500
DTEND;TZID=America/New_York:20150828T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26942
SUMMARY:Probability and Financial Mathematics Seminar - Markov processes in
a random environment
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
Markov processes in a random environment\nSpeaker: Yuri Suhov\, Penn State
University\nAbstract: Abstract. We propose a construction of a Markov pro
cess (MP) in a (Markovian) random environment. (I am not 100 percent sure:
may be some special cases/elements of this constructions can be found in
the existing literature.) A feature of this construction is that it allows
an invariant measure (IM) which is naturally built from IMs for the basic
MPs and IMs for the MP (or (MPs)) describing the dynamics of state of env
ironment (SE). In general tems\, the generator of the combined process is
obtained as a sum of generators for components (with non-commuting summand
s). This construction gives quite spectacular results for some interesting
examples: Jackson network\, simple exclusion\, Ornstein--Uhlenbeck. (The
latter is related to the concept of stochastic volatility in Math Finance.
)\n\nIn the course of presentation\, I will not assume any special knowled
ge from the theory of Markov processes or their applications. However\, ex
posure to basic probabilistic concepts would make understanding easier.
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20150904T153500
DTEND;TZID=America/New_York:20150904T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26945
SUMMARY:Probability and Financial Mathematics Seminar - Algorithmic Stabili
ty in Adaptive Data Analysis
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
Algorithmic Stability in Adaptive Data Analysis\nSpeaker: Adam Smith\, PSU
: Computer Science and Engineering Department\nAbstract: Adaptivity is an
important feature of modern data analysis—often\, the\nchoice of questio
ns asked about a dataset depends on previous\ninteractions with the same d
ataset. Adaptivity can arise in a single\nstudy (say\, when a researcher c
hoses which model to fit based on some\nexploratory data analysis) or\, mo
re subtly\, when data sets are shared\nand re-used across multiple studies
. Unfortunately\, most of the\nstatistical inference theory used in empiri
cal sciences to control\nfalse discovery rates\, and in machine learning t
o avoid overfitting\,\nassumes that the analyses to be performed are selec
ted independently\nof the data. If the set of analyses run is itself a fun
ction of the\ndata\, much of this theory becomes invalid.\n\nSpecifically
\, suppose there is an unknown distribution P and a set of\nn independent
samples x is drawn from P. We seek an algorithm that\,\ngiven x as input\,
“accurately” answers a sequence of adaptively chosen\n“queries” a
bout the unknown distribution P. How many samples n must we\ndraw from the
distribution\, as a function of the type of queries\, the\nnumber of quer
ies\, and the desired level of accuracy?\n\nIn this work we make two new c
ontributions towards resolving this question:\n1. We give upper bounds on
the number of samples n that are needed to\nanswer "statistical queries" t
hat improve over the bounds in the\nrecent work of Dwork et al. (2015).\n2
. We prove the first upper bounds on the number of samples required\nto an
swer more\ngeneral families of queries. These include arbitrary low-sensit
ivity\nqueries and convex risk minimization queries.\n\nOur algorithms are
based on a connection between generalization error\nand a distributional
stability condition on inference algorithms\,\ncalled "differential privac
y".\n\nThe talk will be self-contained.\n\nBased on joint work with Raef B
assily\, Kobbi Nissim\, Thomas Steinke\,\nUri Stemmer and Jon Ullman. http
://arxiv.org/abs/1503.04843\n\nFor some nontechnical background reading\,
see\nGelman and Lokem\, "The Garden of Forking Paths".\nhttp://www.stat.co
lumbia.edu/~gelman/research/unpublished/p_hacking.pdf
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20150911T153500
DTEND;TZID=America/New_York:20150911T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26948
SUMMARY:Probability and Financial Mathematics Seminar - Parametrized circul
ar unitary ensembles.
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
Parametrized circular unitary ensembles.\nSpeaker: Manfred Denker\, PSU\nA
bstract: We discuss extensions of circular unitary ensembles from random m
atrix theory\, which are based on Dyson's classical work in 1962. These mo
dels can be treated using Toeplitz determinants and Szego's determinatal i
dendity for integrable functions on the n-dimensional torus. We show a cet
ral limit heorem for the sufficient statistics defining the models and the
ir associated aproximate maximum likelihood function.
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20150918T153500
DTEND;TZID=America/New_York:20150918T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26951
SUMMARY:Probability and Financial Mathematics Seminar - Ergodic control of
multiclass parallel server systems in the Halfin-Whitt regime
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
Ergodic control of multiclass parallel server systems in the Halfin-Whitt
regime\nSpeaker: Guodong (Gordon) Pang\, Penn State University\nAbstract:
We study the optimal scheduling problem for Markovian multiclass networks
under the long-run average (ergodic) cost criteria in the Halfin-Whitt reg
ime. The arrival processes are Poisson\, service times are exponentially d
istributed with class and pool dependent rates\, and customer patience tim
es are class dependent. For systems with multiple server pools\, we consid
er two formulations: (i) both queueing and idleness costs are minimized\,
and (ii) the queueing cost is minimized while a constraint is imposed upon
the idleness of all server pools. The optimal solution of the scheduling
problem is approximated by that of the ergodic diffusion control in the li
mit via the HJB equations. We introduce a broad class of ergodic diffusion
control problems for diffusions\, which includes the limiting diffusions
for a large class of multiclass multi-pool queueing networks. We also prov
e the asymptotic convergence of the values for the multiclass queueing con
trol problems to the value of the associated ergodic diffusion control pro
blem. The proof relies on an approximation method by spatial truncations f
or the ergodic control of diffusions\, where the Markov policies follow a
fixed priority policy outside a compact set.
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20150925T153500
DTEND;TZID=America/New_York:20150925T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26954
SUMMARY:Probability and Financial Mathematics Seminar - Joint Geometry and
Probability: Random nodal portraits: recent progress and open questions
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
Joint Geometry and Probability: Random nodal portraits: recent progress an
d open questions\nSpeaker: Fedor Nazarov\, Kent University\nAbstract: This
is an overview of my joint work with Mikhail Sodin\non the number of the
nodal components of Gaussian random functions.\nI'll try to address both t
he questions and the techniques (which\,\nsurprisingly\, range from measur
e concentration to the ergodic theorem).
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20151002T153500
DTEND;TZID=America/New_York:20151002T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=29835
SUMMARY:Probability and Financial Mathematics Seminar - Ergodic control of
multiclass parallel server systems in the Halfin-Whitt regime Part II
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
Ergodic control of multiclass parallel server systems in the Halfin-Whitt
regime Part II\nSpeaker: Guodong (Gordon) Pang\, Penn State University\nAb
stract: This is the second part of my previous talk. We study the optimal
scheduling problem for Markovian multiclass networks under the long-run av
erage (ergodic) cost criteria in the Halfin-Whitt regime. The arrival proc
esses are Poisson\, service times are exponentially distributed with class
and pool dependent rates\, and customer patience times are class dependen
t. For systems with multiple server pools\, we consider two formulations:
(i) both queueing and idleness costs are minimized\, and (ii) the queueing
cost is minimized while a constraint is imposed upon the idleness of all
server pools. The optimal solution of the scheduling problem is approximat
ed by that of the ergodic diffusion control in the limit via the HJB equat
ions. We introduce a broad class of ergodic diffusion control problems for
diffusions\, which includes the limiting diffusions for a large class of
multiclass multi-pool queueing networks. We also prove the asymptotic conv
ergence of the values for the multiclass queueing control problems to the
value of the associated ergodic diffusion control problem. The proof relie
s on an approximation method by spatial truncations for the ergodic contro
l of diffusions\, where the Markov policies follow a fixed priority policy
outside a compact set.
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20151009T153500
DTEND;TZID=America/New_York:20151009T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26960
SUMMARY:Probability and Financial Mathematics Seminar - Diffusion on social
networks
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
Diffusion on social networks\nSpeaker: Kalyan Chatterjee\, Penn State Econ
omics\nAbstract: This presentation will discuss two papers in which indivi
dual agents interact on given social networks. In the first model\, the ag
ents behave according to simple rules or heuristics in reacting to their n
eighbours’ choices and outcomes. In the second\, players rationally deci
de whether to pass on messages they receive. The first model studies the e
xtent of diffusion on the line\; the second also investigates characterist
ics of the network that could lead to desirable outcomes.
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20151016T153500
DTEND;TZID=America/New_York:20151016T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26963
SUMMARY:Probability and Financial Mathematics Seminar - TBA
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
TBA
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20151023T153500
DTEND;TZID=America/New_York:20151023T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26966
SUMMARY:Probability and Financial Mathematics Seminar - TBA
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
TBA
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20151030T153500
DTEND;TZID=America/New_York:20151030T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26969
SUMMARY:Probability and Financial Mathematics Seminar - Which Firm Characte
ristics Predict Stock Returns and When? A Hierarchical Bayesian Variable S
election Approach.
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
Which Firm Characteristics Predict Stock Returns and When? A Hierarchical
Bayesian Variable Selection Approach.\nSpeaker: John Liechty\, Penn State
\, Marketing and Statistics\nAbstract: We develop and apply an enhanced Ba
yesian variable selection methodology (a hierarchical variable dimension\,
variable selection approach) to the problem of choosing which firm charac
teristics predict the cross-section of returns and when. We find that almo
st all firm characteristics matter\, but that they do not matter all of th
e time. In stark contrast to the findings in prior research\, momentum has
a low impact on predictions as it is rarely selected (on the order of 6
\\% of the time) and when it is selected the corresponding slope is relati
vely small\, except during the Internet bubble. In contrast\, size has a s
trong impact both in terms of the percentage of months it is included and
the size of the corresponding slope\, which changes sign and varies dramat
ically over time. Macro variables used to predict when and how much a part
icular firm characteristic predicts returns are only partially supported b
y the data\; and in particular\, macro-sentiment related variables have no
effect. We do find that credit market macro variables tend to be the prim
ary drivers of how much firm characteristics predict returns and that busi
ness cycle macro variables drive when firm characteristics predict returns
\; with firm characteristics more likely to predict returns during periods
of market expansion. We discuss other findings and the implications of ou
r results for research studying stock returns and for asset management.
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20151106T153500
DTEND;TZID=America/New_York:20151106T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26972
SUMMARY:Probability and Financial Mathematics Seminar - TBA
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
TBA
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20151113T153500
DTEND;TZID=America/New_York:20151113T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26975
SUMMARY:Probability and Financial Mathematics Seminar - TBA
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
TBA
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20151120T153500
DTEND;TZID=America/New_York:20151120T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26978
SUMMARY:Probability and Financial Mathematics Seminar - TBA
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
TBA
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20151127T153500
DTEND;TZID=America/New_York:20151127T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26981
SUMMARY:Probability and Financial Mathematics Seminar - TBA
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
TBA
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20151204T153500
DTEND;TZID=America/New_York:20151204T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26984
SUMMARY:Probability and Financial Mathematics Seminar - TBA
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
TBA
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20151211T153500
DTEND;TZID=America/New_York:20151211T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26987
SUMMARY:Probability and Financial Mathematics Seminar - TBA
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
TBA
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20151218T153500
DTEND;TZID=America/New_York:20151218T163500
LOCATION:MB106
URL:http://www.math.psu.edu/seminars/meeting.php?id=26990
SUMMARY:Probability and Financial Mathematics Seminar - TBA
DESCRIPTION:Seminar: Probability and Financial Mathematics Seminar\nTitle:
TBA
END:VEVENT
END:VCALENDAR