BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Quantification of model inadequacy within high-dimensional Bayesia
n inverse problems
DTSTART;VALUE=DATE-TIME:20161108T151500Z
DTEND;VALUE=DATE-TIME:20161108T154500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2264@indico.ph.tum.de
DESCRIPTION:Speakers: Isabell Franck (TUM)\nWhile calibration can almost a
lways been archived it becomes problematic if the underlying model is inco
rrect\, which will lead to wrong predictions and interpretations. Traditio
nal approaches use an additional regression model (e.g. GP) added to the m
odel output or within a submodel to account for an underlying model error.
This can either violate physical constraints and/or is infeasible in high
dimensions. In this work we unfold conservation and constitutive laws to
estimate model discrepancies accurately and use Variational Bayes to decre
ase computational costs. We investigate this problem within a high-dimensi
onal inverse problem from solid mechanics where an identification of the m
echanical properties can lead to noninvasive\, medical diagnosis.\n\nhttps
://indico.ph.tum.de/event/3597/contributions/2264/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2264/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Operator Calculus for Information Field Theory
DTSTART;VALUE=DATE-TIME:20161108T154500Z
DTEND;VALUE=DATE-TIME:20161108T161500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2263@indico.ph.tum.de
DESCRIPTION:Speakers: Reimar Leike (MPA)\nSignal inference problems with n
on-Gaussian posteriors are very hard to tackle. Through using the concept
of Gibbs free energy these problems can be rephrased as Gaussian problems
for the price of computing expectation values of various functions with re
spect to a Gaussian distribution. We present a new way of translating thes
e expectation values to the language of operators which allows us to simpl
ify many calculations\, especially calculations that arose from log-normal
priors which are the natural priors for signals that vary over many order
s of magnitude.\n\nhttps://indico.ph.tum.de/event/3597/contributions/2263/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2263/
END:VEVENT
BEGIN:VEVENT
SUMMARY:NIFTy and D2O: A framework for numerical IFT implementations
DTSTART;VALUE=DATE-TIME:20161108T161500Z
DTEND;VALUE=DATE-TIME:20161108T164500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2262@indico.ph.tum.de
DESCRIPTION:Speakers: Theo Steininger (MPA)\nNIFTy\, “Numerical Informat
ion Field Theory”\, is a versatile library designed to enable the develo
pment of signal inference algorithms that operate regardless of the underl
ying spatial grid and its resolution. Its object-oriented framework is wri
tten in Python\, although it accesses libraries written in Cython\, C++\,
and C for efficiency. NIFTy offers a toolkit that abstracts discretized re
presentations of continuous spaces\, fields in these spaces\, and operator
s acting on fields into classes. In order to utilize the power of high-per
formance computing clusters NIFTy is built on D2O\, a Python module for cl
uster-distributed multi-dimensional numerical arrays. An overview will be
given.\n\nhttps://indico.ph.tum.de/event/3597/contributions/2262/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2262/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Population inference in Astronomy - Demographics from limited samp
les\, luminosity functions\, hierarchical models
DTSTART;VALUE=DATE-TIME:20161109T103000Z
DTEND;VALUE=DATE-TIME:20161109T110000Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2261@indico.ph.tum.de
DESCRIPTION:Speakers: Johannes Buchner (MPE)\nHierarchical models allow to
infer distributions of an underlying population from a sample of detected
objects with uncertain measurements. Since decades\, astronomers have emp
loyed what is now popular as hierarchical (Bayesian) inference\, including
the consistent handling of selection biases. Yet these powerful methods a
re under-used today when dealing with samples and their uncertainties. I w
ill give an introduction and demonstrate that the method is simple to appl
y for a wide range of problems\, including deriving the intrinsic luminosi
ty distribution of a flux-limited sample\, estimation of population proper
ties and distinction between physical models.\n\nhttps://indico.ph.tum.de/
event/3597/contributions/2261/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2261/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Modeling star and planet formation: how to infer the physics from
observations?
DTSTART;VALUE=DATE-TIME:20161109T083000Z
DTEND;VALUE=DATE-TIME:20161109T090000Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2260@indico.ph.tum.de
DESCRIPTION:Speakers: Leonardo Testi (ESO)\nUnderstanding the physics that
governs the star and planet formation process is one of the major challen
ges of modern astrophysics. These are key processes that govern the cyclin
g of diffuse matter into stars and the formation of planetary systems. The
key difficulties that we face as observational astrophysicists are that w
e are not able to run controlled experiments. We do observe the possible i
nitial conditions\, the processes at work\, and their final outcome\, but
not as a well defined sequential experiment\, as\, in most cases\, we can
only make educated guesses on the past history and future outcome of any o
bject we observe.\nLeonardo will present two approaches that can be follow
ed to understand the underlying physics in the star and plant formation pr
ocess: simplified modeling of large samples of objects\, to identify the m
ain physical properties\, and the comparison of detailed numerical simulat
ions with observed objects. This latter approach offer the possibility of
running controlled (numerical) experiments\, but the challenge is to compa
re these with real observations.\n\nhttps://indico.ph.tum.de/event/3597/co
ntributions/2260/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2260/
END:VEVENT
BEGIN:VEVENT
SUMMARY:SOMBI: Bayesian identification of parameter relations in unstructu
red cosmological data
DTSTART;VALUE=DATE-TIME:20161108T130000Z
DTEND;VALUE=DATE-TIME:20161108T133000Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2259@indico.ph.tum.de
DESCRIPTION:Speakers: Philipp Frank (LMU)\nGenerally\, cosmological or ast
ronomical observations are generated by a combination of different physica
l effects. Separating these observations into different subgroups\, which
permit to study individual aspects of the corresponding physical theory\,
is a particularly challenging task in regimes of huge datasets or in high
dimensions where manual clustering of data is not feasible. As a response
to this problem we present SOMBI\, a Bayesian inference approach to search
for data clusters and relations between observed parameters without human
intervention. SOMBI aims to automatically identify relations between diff
erent observed parameters by first identifying data clusters in high-dimen
sional datasets via the self organizing map neural network algorithm. Para
meter relations are then revealed by means of a Bayesian inference within
respective identified data clusters.\n\nhttps://indico.ph.tum.de/event/359
7/contributions/2259/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2259/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Model comparison challenges in ATLAS
DTSTART;VALUE=DATE-TIME:20161107T103000Z
DTEND;VALUE=DATE-TIME:20161107T110000Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2258@indico.ph.tum.de
DESCRIPTION:Speakers: Jeanette Lorenz (LMU)\nThe ATLAS experiment searches
for new physics in the enormous amount of data recorded in proton-proton
collisions delivered by the Large Hadron Collider (LHC)\, and measures the
properties of the Standard Model of Particle Physics to increasing higher
precision. The challenges in a typical search for new physics are numerou
s\, but one key aspect is the construction of the best possible statistica
l model to describe the data most precisely which allows to test for any k
ind of new physics. This talk will describe the statistical methods and mo
dels used by the ATLAS experiment\, with a particular emphasis on how to s
et limits on new physics or how to claim discovery.\n\nhttps://indico.ph.t
um.de/event/3597/contributions/2258/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2258/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Machine learning methods in Astrophysics
DTSTART;VALUE=DATE-TIME:20161108T080000Z
DTEND;VALUE=DATE-TIME:20161108T084500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2257@indico.ph.tum.de
DESCRIPTION:Speakers: Giuseppe Longo (University Federico II\, Napoli and
Caltech\, USA)\nhttps://indico.ph.tum.de/event/3597/contributions/2257/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2257/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Sampling from a Gaussian in high dimensions
DTSTART;VALUE=DATE-TIME:20161109T134500Z
DTEND;VALUE=DATE-TIME:20161109T141500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2256@indico.ph.tum.de
DESCRIPTION:Speakers: Frederic Beaujean (LMU)\nSamples from a Gaussian for
m the basis of many more sophisticated algorithms. While simple in low num
ber of dimensions\, it turns out to be very challenging in millions of dim
ensions when matrix inversions or factorizations have to be avoided becaus
e of poor scaling. Fred will present a comparison of reflective slice samp
ling and Hamiltonian Monte Carlo\, both relying on the gradient of the tar
get density.\n\nhttps://indico.ph.tum.de/event/3597/contributions/2256/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2256/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Dynamic System Classifier
DTSTART;VALUE=DATE-TIME:20161107T100000Z
DTEND;VALUE=DATE-TIME:20161107T103000Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2255@indico.ph.tum.de
DESCRIPTION:Speakers: Daniel Pumpe (MPA)\nStochastic differential equation
s describe well many physical\, biological and sociological systems\, desp
ite the simplification often made in their derivation. Here the usage of s
imple stochastic differential equations to characterize and classify compl
ex dynamical systems is proposed within a Bayesian framework. To this end\
, we develop a dynamic system classifier (DSC). The DSC first abstracts tr
aining data of a system in terms of time dependent coefficients of the des
criptive stochastic differential equation. Thereby the DSC identifies uniq
ue correlation structures within the training data. For definite- ness we
restrict the presentation of DSC to oscillation processes with a time depe
ndent frequency ω(t) and damping factor γ(t). Although real systems migh
t be more com- plex\, this simple oscillator captures many characteristic
features. The ω and γ timelines represent the abstract system characteri
zation and permit the construction of efficient signal classifiers. Numeri
cal experiments show that such classifiers perform well even in the low si
gnal-to-noise regime.\n\nhttps://indico.ph.tum.de/event/3597/contributions
/2255/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2255/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Bayesian Component Separation
DTSTART;VALUE=DATE-TIME:20161107T144500Z
DTEND;VALUE=DATE-TIME:20161107T151500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2254@indico.ph.tum.de
DESCRIPTION:Speakers: Jakob Knollmüller (MPA)\nhttps://indico.ph.tum.de/e
vent/3597/contributions/2254/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2254/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Overview of Model Comparison and Testing Approaches
DTSTART;VALUE=DATE-TIME:20161107T080000Z
DTEND;VALUE=DATE-TIME:20161107T084500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2253@indico.ph.tum.de
DESCRIPTION:Speakers: Allen Caldwell (MPP)\nModel testing and model compar
ison are performed in a number of ways in the sciences\, and there is no c
onsensus concerning best practice. The conceptual basis for different appr
oaches will be presented. Model testing in frequentist and Bayesian style
analysis will be discussed\, and model comparison and model selection usin
g Bayes factors\, p-values and frequentist tests will be reviewed and comm
ented.\n\nhttps://indico.ph.tum.de/event/3597/contributions/2253/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2253/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Robust Parameter estimation
DTSTART;VALUE=DATE-TIME:20161109T141500Z
DTEND;VALUE=DATE-TIME:20161109T144500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2252@indico.ph.tum.de
DESCRIPTION:Speakers: Udo von Toussaint (IPP)\nData collections\, like ast
ronomical surveys or sets of simulation results generally contain entries
with limited information about their accuracy. In addition\, the ubiquitou
s assumption of Gaussian uncertainties is often not adequate. We review th
e nature of this problem and present recent methodological developments in
the area of robust estimation. In particular recent developments in the a
pplication of L1-regression techniques to multidimensional gridded data se
ts are discussed.\n\nhttps://indico.ph.tum.de/event/3597/contributions/225
2/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2252/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Scalable Scientific Data Visualization
DTSTART;VALUE=DATE-TIME:20161107T130000Z
DTEND;VALUE=DATE-TIME:20161107T134500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2251@indico.ph.tum.de
DESCRIPTION:Speakers: Rüdiger Westermann (TUM)\nIn this talk a number of
recent developments in the area of large data visualization at the Chair f
or Computer Graphics and Visualization at TUM will be discussed. Some of t
he results of this research will be presented\, which have been achieved i
n collaboration with astrophysicists and meteorologists. The focus of the
talk will be on scalability issues with respect to both the increasing amo
unt of data and the increasing complexity of this data. Recent approaches
for data compression and feature extraction\, as well as approaches for vi
sualizing the uncertainty that is present in ensembles of fields will be s
hown. In addition\, the use of parallel graphics hardware to achieve inter
activity will be demonstrated.\n\nhttps://indico.ph.tum.de/event/3597/cont
ributions/2251/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2251/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Mixture models in high dimensions
DTSTART;VALUE=DATE-TIME:20161108T104500Z
DTEND;VALUE=DATE-TIME:20161108T111500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2250@indico.ph.tum.de
DESCRIPTION:Speakers: Maksim Greiner (TUM)\nMaksim will talk about density
estimation using mixture models\, more specifically Gaussian and categori
cal mixture models. He will address their performance with increasing numb
ers of parameters (1000 parameters and more) and with complex non-linear d
ependencies between the parameters. He will also talk about conditional sa
mpling from the trained mixture model.\n\nhttps://indico.ph.tum.de/event/3
597/contributions/2250/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2250/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Challenges of atmospheric data assimilation
DTSTART;VALUE=DATE-TIME:20161107T134500Z
DTEND;VALUE=DATE-TIME:20161107T141500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2249@indico.ph.tum.de
DESCRIPTION:Speakers: Tijana Janjic Pfander (LMU Meteorology)\nIn this tal
k\, we present the mechanisms of the data assimilation algorithms on examp
les that range from toy models to atmospheric applications. We focus on th
e ensemble Kalman filter algorithm to estimate the atmospheric state as we
ll as its necessary modifications for our application. We argue that relax
ing underlying assumptions of the data assimilation algorithms might be po
ssible by improving the link between the data assimilation and the model.
For example\, the stronger connection can be established by constraining t
he analysis with imposing conservation laws and other physical constraints
. Besides the inclusion of constraints in order to obtain more physically
based solution that is consistent with both the nature and the prediction
model\, the problem of the representativeness error (mismatch of scales an
d processes present in observation and model) and model error are discusse
d.\n\nhttps://indico.ph.tum.de/event/3597/contributions/2249/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2249/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Rosat-2RXS counterparts using Nway – An accurate algorithm to pa
ir sources simultaneously between N catalogs
DTSTART;VALUE=DATE-TIME:20161108T094500Z
DTEND;VALUE=DATE-TIME:20161108T101500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2248@indico.ph.tum.de
DESCRIPTION:Speakers: Mara Salvato (MPE)\nThe increasing number of surveys
available at any wavelength is allowing the construction of Spectral Ener
gy Distribution (SED) for any kind of astrophysical object. However\, a) d
ifferent surveys/instruments\, in particular at X-ray\, UV and MIR wavelen
gth\, have different positional accuracy and resolution and b) the surveys
depth do not match each other and depending on redshift and SED\, a give
n source might or might not be detected at a certain wavelength. All this
makes the pairing of sources among catalogs not trivial\, specially in cro
wded fields. In order to overcome this issue\, we propose a new algorithm
that combine the best of Bayesian and frequentist methods but that can be
used as the common Likelihood Ratio (LR) technique in the simplest of the
applications. In this talk Mara will introduce the code and how it has bee
n used for finding the ALLWISE counterparts to the X-ray ROSAT All-sky sur
vey.\n\nhttps://indico.ph.tum.de/event/3597/contributions/2248/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2248/
END:VEVENT
BEGIN:VEVENT
SUMMARY:New Smart Data Services of the Digital Future
DTSTART;VALUE=DATE-TIME:20161108T101500Z
DTEND;VALUE=DATE-TIME:20161108T104500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2247@indico.ph.tum.de
DESCRIPTION:Speakers: René Fassbender (OmegaLambdaTec)\nThe ongoing digit
al transformation produces massive amounts of new data at every instant th
at may be used as the basis for the development of novel Smart Data Servic
es with an increasing importance for our everyday life. In this talk René
will review the status of data driven applications and will give an outlo
ok on Smart Data innovations that will be possible in the next few years.
He will discuss examples of current data challenges to be solved in the fi
elds of Smart Energy\, Smart Mobility\, Smart City\, Smart Factory as well
as consumer data innovation.\n\nhttps://indico.ph.tum.de/event/3597/contr
ibutions/2247/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2247/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Quantifying Tensions between Independent Data Sets
DTSTART;VALUE=DATE-TIME:20161107T084500Z
DTEND;VALUE=DATE-TIME:20161107T091500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2246@indico.ph.tum.de
DESCRIPTION:Speakers: Sebastian Grandis (LMU)\nModern Cosmology is blessed
by a wealth of high-precision data sets\, putting independent constraints
on the underlying cosmological model. In this context\, an important test
for every model is the consistency and agreement of the constraints deriv
ed from the different independent measurements. Given the heterogeneity of
cosmological data sets\, this comparison is best performed in the space o
f model parameters. We present here a recently development measure of data
set consistency\, the ‘Surprise’\, derived from the information theor
y. After comparing it to other proposed measures of data set agreement\, w
e show how the Surprise can be estimated from samples of the prior and pos
terior distributions. Furthermore\, we present different applications of t
he Surprise in cosmological context.\n\nhttps://indico.ph.tum.de/event/359
7/contributions/2246/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2246/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Machine learning in Cosmology
DTSTART;VALUE=DATE-TIME:20161108T084500Z
DTEND;VALUE=DATE-TIME:20161108T091500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2245@indico.ph.tum.de
DESCRIPTION:Speakers: Ben Hoyle (LMU)\nBen will provide a brief introducti
on to machine learning\, and discuss within which regimes it is a suitable
statistical tool. He will highlight recent uses of machine learning in th
e astrophysics and cosmology literature\, and describe some recent project
s for which he has found machine learning to be a competitive tool\, inclu
ding star galaxy separation\, optimised target selection\, and photometric
redshifts.\n\nhttps://indico.ph.tum.de/event/3597/contributions/2245/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2245/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Compressive Computation
DTSTART;VALUE=DATE-TIME:20161108T133000Z
DTEND;VALUE=DATE-TIME:20161108T141500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2244@indico.ph.tum.de
DESCRIPTION:Speakers: John Skilling (MaxEnt Data Consultant Ltd.)\nAsymmet
ry between prior proposals and posterior conclusions lies at the heart of
inference. Passage between the two is generally irreversible. The prior su
pports the posterior\, but *not* the other way round. Prior possibilities
are eliminated when modulated by a zero likelihood factor\, and cannot be
recovered because dividing by that zero is impossible. Accordingly\, the t
ime-honoured techniques of reversible detailed balance do not apply. Inste
ad\, we need a one-way algorithm to do systematic compression. The standar
d technique of “simulated annealing” introduces likelihood modulation
gradually through fractional powers\, which fails because any fractional p
ower of zero is still zero. That shows up as failure to compute phase chan
ges. Compression is best accomplished iteratively by elimination of succes
sive layers of low likelihood\, which amounts to programming Lebesgue inte
gration. This direct and general technique is “nested sampling”.\n\nht
tps://indico.ph.tum.de/event/3597/contributions/2244/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2244/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Sparse Optimal Control in Measure Spaces
DTSTART;VALUE=DATE-TIME:20161109T151500Z
DTEND;VALUE=DATE-TIME:20161109T154500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2243@indico.ph.tum.de
DESCRIPTION:Speakers: Boris Vexler (TUM)\nIn this talk we consider optimal
control problem governed by elliptic and parabolic equations\, where the
control variable lies in a measure space. Such formulations lead to a spar
se structure of the optimal control\, which provides among other things an
elegant way to attack problems of optimal actuator or sensor placement as
well as point source identification problems. We discuss the functional a
nalytic settings of such problems and the regularity issues of the optimal
solutions. Moreover\, we present a discretization concept and discuss err
or estimates for the discretization error.\n\nhttps://indico.ph.tum.de/eve
nt/3597/contributions/2243/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2243/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Physical Modelling of X-Shooter data
DTSTART;VALUE=DATE-TIME:20161109T090000Z
DTEND;VALUE=DATE-TIME:20161109T093000Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2242@indico.ph.tum.de
DESCRIPTION:Speakers: Wolfgang Kerzendorf (ESO)\nOne of the main steps in
analysis and calibration of optical spectroscopy is to remove the instrume
nt as well as the atmosphere signature. In particular\, it means generatin
g a one dimensional spectrum (wavelength\, intensity) from a 2D CCD frame.
\nWolfgang has developed a model that allows a good reconstruction of thi
s 2D Frame and is working on the fitting routines. In this talk he will sh
ow how combining linear least squares with non-linear least squares is a g
ood approach to combat the large number of parameters in a novel technique
. Wolfgang will conclude by showing first results and discussion about the
broader applications of this technique.\n\nhttps://indico.ph.tum.de/event
/3597/contributions/2242/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2242/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Performance enhancement via surrogate minimization
DTSTART;VALUE=DATE-TIME:20161109T154500Z
DTEND;VALUE=DATE-TIME:20161109T161500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2241@indico.ph.tum.de
DESCRIPTION:Speakers: Fabrizia Guglielmetti (MPA)\nFinding the Maximum a P
osteriori solution is a numerically challenging problem\, especially when
estimating an expensive objective function defined in an high-dimensional
design domain. We propose to use Kriging surrogates to speed up optimizati
on schemes\, like steepest descent. Surrogate models are built and incorpo
rated in a sequential optimization strategy. Results are presented with ap
plication on astronomical images\, showing the proposed method can effecti
vely search the global optimum.\n\nhttps://indico.ph.tum.de/event/3597/con
tributions/2241/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2241/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Large scale Bayesian inference in cosmology
DTSTART;VALUE=DATE-TIME:20161109T080000Z
DTEND;VALUE=DATE-TIME:20161109T083000Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2240@indico.ph.tum.de
DESCRIPTION:Speakers: Jens Jasche (TUM)\nPresently proposed and designed f
uture cosmological probes and surveys permit us to anticipate the upcoming
avalanche of cosmological information during the next decades. The increa
se of valuable observations needs to be accompanied with the development o
f efficient and accurate information processing technology in order to ana
lyse and interpret this data. The analysis of the structure and evolution
of our inhomogeneous Universe therefore requires to solve non-linear stati
stical inference problems in very high dimensional parameter spaces\, invo
lving on the order of 10^7 or more parameters. In this talk Jens will addr
ess the problem of high dimensional Bayesian inference from cosmological d
ata sets via the recently proposed BORG algorithm. This method couples an
approximate model of structure formation to an Hybrid Monte Carlo algorith
m providing a fully probabilistic\, physical model of the non-linearly evo
lved density field as probed by galaxy surveys. Besides highly accurate an
d detailed measurements of three dimensional density and velocity fields\,
this methodology also infers plausible dynamic formation histories for th
e observed large scale structure.\n\nhttps://indico.ph.tum.de/event/3597/c
ontributions/2240/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2240/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Information field theory
DTSTART;VALUE=DATE-TIME:20161108T141500Z
DTEND;VALUE=DATE-TIME:20161108T144500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2239@indico.ph.tum.de
DESCRIPTION:Speakers: Torsten ENSSLIN (MPA)\nInformation field theory (IFT
) describes probabilistic image reconstruction from incomplete and noisy d
ata. Based on field theoretical concepts IFT provides optimal methods to g
enerate images exploiting all available information. Applications in astro
physics are galactic tomography\, gamma- and radio- astronomical imaging\,
and the analysis of cosmic microwave background data.\n\nhttps://indico.p
h.tum.de/event/3597/contributions/2239/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2239/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Adaptive sparse grids for high-dimensional machine learning
DTSTART;VALUE=DATE-TIME:20161109T130000Z
DTEND;VALUE=DATE-TIME:20161109T134500Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2238@indico.ph.tum.de
DESCRIPTION:Speakers: Valeriy Khakhutskyy (TUM)\nhttps://indico.ph.tum.de/
event/3597/contributions/2238/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2238/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Modeling gene expression from genetic data
DTSTART;VALUE=DATE-TIME:20161109T100000Z
DTEND;VALUE=DATE-TIME:20161109T103000Z
DTSTAMP;VALUE=DATE-TIME:20180225T100220Z
UID:indico-contribution-3597-2236@indico.ph.tum.de
DESCRIPTION:Speakers: Julien Gagneur (TUM)\nThe cells of nearly all cell t
ypes of your body contain the same copy of your genome. Why do they do so
different things? Different cells read different parts of the genome. For
each of the 22\,000 genes of your genome\, the amount of RNA and protein m
olecules that a cell makes depends on the cell type and on stimulations fr
om the cellular environment. In turn\, this regulatory program is encoded
in the genome\, within and also between the genes. Understanding the genet
ic regulatory code and how errors in the regulatory program can lead to di
seases is the research topic of Julien’s lab. In this talk\, he will pre
sent statistical and machine leaning models which\, by integrating large-s
cale genetic and molecular datasets\, help deciphering the regulatory code
and how these models allow predicting quantitatively effects of genetic m
utations on gene regulation. He will finish by showing how such integrativ
e analyses can help pinpointing the genetic defects of patients with rare
diseases.\n\nhttps://indico.ph.tum.de/event/3597/contributions/2236/
LOCATION:Exzellenzzentrum / IGSSE. Seminar room\, ground floor (5530.EG.00
3) Seminar room\, ground floor (5530.EG.003)
URL:https://indico.ph.tum.de/event/3597/contributions/2236/
END:VEVENT
END:VCALENDAR