BEGIN:VCALENDAR
PRODID:-//Campus Event Calendar//EN
X-WR-CALNAME: Campus Event Calendar
X-WR-TIMEZONE:(UTC+01:00) Amsterdam, Berlin, Bern, Rome, Stockholm, Vienna
VERSION:2.0
METHOD:PUBLISH
BEGIN:VTIMEZONE
TZID:Europe/Berlin
X-LIC-LOCATION:Europe/Berlin
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700329T020000
RRULE:FREQ=YEARLY;INTERVAL=1;BYDAY=-1SU;BYMONTH=3
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701025T030000
RRULE:FREQ=YEARLY;INTERVAL=1;BYDAY=-1SU;BYMONTH=10
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=Europe/Berlin:20190122T100000
DURATION:PT1H0M0S
UID:E2F80186AF127955C125838400394C3C
DTSTAMP:20190116T112550
SUMMARY:Florian Frohn (RWTH Aachen) talks about "Automated Complexity
Analysis of Rewrite Systems"
LOCATION:Kaiserslautern building G26, room 111 / simultaneous videocas
t to Saarbrücken building E1 5, room 029
CATEGORIES:SWS Colloquium [AG Audience]
CONTACT:Susanne Girard
DESCRIPTION:Many approaches to analyze the complexity of programs auto
matically are based \non transformations into integer or term rewrite
systems. However, \nstate-of-the-art tools that analyze the worst-case
complexity of rewrite \nsystems are restricted to the inference of up
per bounds. In this talk, the \nfirst techniques for the inference of
lower bounds on the worst-case complexity \nof integer and term rewrit
e systems are introduced. While upper bounds can \nprove the absence o
f performance-critical bugs, lower bounds can be used to \nfind them.\
n\nFor term rewriting, the power of the presented technique gives rise
to the \nquestion whether the existence of a non-constant lower bound
is decidable. \nThus, the corresponding decidability results are also
discussed in this talk.\n\nFinally, to see the practical value of com
plexity analysis techniques for \nrewrite systems, we will have a glim
pse at the transformation from Java \nprograms to integer rewrite syst
ems that is implemented in the tool AProVE.
TRANSP:OPAQUE
CLASS:PUBLIC
URL:http://domino.mpi-inf.mpg.de/internet/events.nsf/all/20190122-1000
-111
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Berlin:20190122T130000
DURATION:PT0H30M0S
UID:3CCE6A9901D3E2F7C1258382002A6C70
DTSTAMP:20190114T084322
SUMMARY:Stefano Leucci (MPI-INF - D1) talks about "Optimal Sorting wit
h Persistent Comparison Errors"
LOCATION:Saarbrücken building E1 4, room 024
CATEGORIES:AG1 Mittagsseminar (own work) [AG Audience]
CONTACT:Nitin Saurabh
DESCRIPTION:We consider the problem of sorting n elements in the case
of persistent \ncomparison errors. In this problem, each comparison b
etween two elements can \nbe wrong with some fixed (small) probabilit
y p, and comparisons cannot be \nrepeated (Braverman and Mossel, SODA'
08). \nSorting perfectly in this model is impossible, and the objectiv
e is to minimize \nthe dislocation of each element in the output seque
nce, that is, the difference \nbetween its true rank and its position.
Existing lower bounds for this problem \nshow that no algorithm can g
uarantee, with high probability, maximum \ndislocation and total dislo
cation better than Omega(log n) and Omega(n), \nrespectively, regardle
ss of its running time.\nWe present the first O(n log n)-time sorting
algorithm that guarantees both \nO(log n) maximum dislocation and O(n)
total dislocation with high probability. \n This settles the time com
plexity of this problem and shows that \ncomparison errors do not make
the problem computationally more difficult: a \nsequence with the bes
t possible dislocation can be obtained in O(n log n) time \nand, even
without comparison errors, Omega(n log n) time is necessary to \nguara
ntee such dislocation bounds.
TRANSP:OPAQUE
CLASS:PUBLIC
URL:http://domino.mpi-inf.mpg.de/internet/events.nsf/all/20190122-1300
-024
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Berlin:20190123T161500
DURATION:PT1H0M0S
UID:889510A31F637FCBC1258383004AA804
DTSTAMP:20190115T143526
SUMMARY:Nico Gründel (MPI-INF - D1) talks about "On the Difference Bet
ween Closest, Furthest, and Orthogonal Pairs: Nearly-Linear vs Barely-
Subquadratic Complexity in Computational Geometry"
LOCATION:Saarbrücken building E1 4, room 024
CATEGORIES:MPI-Seminar [AG Audience]
CONTACT:Daniel Vaz
DESCRIPTION:Bla
TRANSP:OPAQUE
CLASS:PUBLIC
URL:http://domino.mpi-inf.mpg.de/internet/events.nsf/all/20190123-1615
-024
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Berlin:20190124T130000
DURATION:PT0H30M0S
UID:082DAC0DE6DE3235C1258386006F730E
DTSTAMP:20190118T211719
SUMMARY:Vasileios Nakos (Harvard University) talks about "Nearly Optim
al Sparse Polynomial Multiplication"
LOCATION:Saarbrücken building E1 4, room 024
CATEGORIES:AG1 Mittagsseminar (own work) [AG Audience]
CONTACT:Karl Bringmann
DESCRIPTION:In the sparse multiplication problem, one is asked to mult
iply two sparse \npolynomials f and g of degree at most n in time that
is proportional to the \nsize of the input plus the size of the outpu
t. The polynomials are given via \nlists of their coefficients F and G
, respectively. Cole and Hariharan (STOC 02) \nhave given a nearly opt
imal algorithm when the coefficients are positive, and \nArnold and No
che (ISSAC 15) devised an algorithm running in time proportional \nto
the "structural sparsity" of the product, i.e. the set supp(F) + supp(
G). \nThe latter algorithm is particularly efficient when there not "t
oo many \ncancellations" of coefficients in the product.\n\nIn this wo
rk we give a clean, nearly optimal algorithm for the sparse \npolynomi
al multiplication problem, resolving an open question posed by Noche \
n(ISSAC 18). Our algorithm is not only more general, but strictly fast
er than \nprevious attempts, and beats FFT whenever the size of the in
put plus the size \nof the output is at most n/log log n.
TRANSP:OPAQUE
CLASS:PUBLIC
URL:http://domino.mpi-inf.mpg.de/internet/events.nsf/all/20190124-1300
-024
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Berlin:20190130T161500
DURATION:PT1H0M0S
UID:60085C4E6C9DB89FC1258383004B5421
DTSTAMP:20190115T144247
SUMMARY:Eunjin Oh (MPI-INF - D1) talks about "Exact Distance Oracles f
or Planar Graphs with Failing Vertices"
LOCATION:Saarbrücken building E1 4, room 024
CATEGORIES:MPI-Seminar [AG Audience]
CONTACT:Daniel Vaz
DESCRIPTION:Bla
TRANSP:OPAQUE
CLASS:PUBLIC
URL:http://domino.mpi-inf.mpg.de/internet/events.nsf/all/20190130-1615
-024
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Berlin:20190204T170000
DURATION:PT1H0M0S
UID:E771A521DB0FC781C1258385004C0FC2
DTSTAMP:20190117T145047
SUMMARY:Bilal Zafar (Max Planck Institute for Software Systems) talks
about "Discrimination in Algorithmic Decision Making: From Principles
to Measures and Mechanisms"
LOCATION:Saarbrücken building E1 5, room 029 / simultaneous videocast
to Kaiserslautern building G26, room 111
CATEGORIES:SWS Student Defense Talks - Thesis Defense [Public Audience]
CONTACT:
DESCRIPTION:The rise of algorithmic decision making in a variety of ap
plications has also \nraised\nconcerns about its potential for discrim
ination against certain social groups. \nHowever,\nincorporating nondi
scrimination goals into the design of algorithmic decision \nmaking\ns
ystems (or, classiﬁers) has proven to be quite challenging. These chal
lenges \narise mainly\ndue to the computational complexities involved
in the process, and the \ninadequacy of\nexisting measures to computat
ionally capture discrimination in various \nsituations. The\ngoal of t
his thesis is to tackle these problems.\n\nFirst, with the aim of inco
rporating existing measures of discrimination \n(namely,\ndisparate tr
eatment and disparate impact) into the design of well-known \nclassiﬁe
rs, we\nintroduce a mechanism of decision boundary covariance, that ca
n be included in \nthe\nformulation of any convex boundary-based class
iﬁer in the form of convex \nconstraints.\nSecond, we propose alternat
ive measures of discrimination. Our ﬁrst proposed \nmeasure,\ndisparat
e mistreatment, is useful in situations when unbiased ground truth \nt
raining data\nis available. The other two measures, preferred treatmen
t and preferred impact, \nare\nuseful in situations when feature and c
lass distributions of different social \ngroups are\nsigniﬁcantly diff
erent, and can additionally help reduce the cost of \nnondiscriminatio
n\n(as compared to the existing measures). We also design mechanisms t
o \nincorporate these\nnew measures into the design of convex boundary
-based classiﬁers.
TRANSP:OPAQUE
CLASS:PUBLIC
URL:http://domino.mpi-inf.mpg.de/internet/events.nsf/all/20190204-1700
-029
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Berlin:20190205T173000
DURATION:PT1H0M0S
UID:B9BFC619214C02E2C1258385004A91CC
DTSTAMP:20190117T143429
SUMMARY:Anjo Vahldiek-Oberwagner (Max Planck Institute for Software Sy
stems) talks about "Techniques to Protect Confidentiality and Integrit
y of Persistant and In-Memory Data"
LOCATION:Saarbrücken building E1 5, room 029 / simultaneous videocast
to Kaiserslautern building G26, room 111
CATEGORIES:SWS Student Defense Talks - Thesis Defense [Public Audience]
CONTACT:
DESCRIPTION:Today computers store and analyze valuable and sensitive d
ata. As a result we \nneed\nto protect this data against confidentiali
ty and integrity violations that can \nresult\nin the illicit release,
loss, or modification of a user’s and an organization’s \nsensitive\n
data such as personal media content or client records. Existing techni
ques \nprotecting\nconfidentiality and integrity lack either efficienc
y or are vulnerable to \nmalicious\nattacks. In this thesis we suggest
techniques, Guardat and ERIM, to efficiently \nand\nrobustly protect
persistent and in-memory data.\n \nTo protect the confidentiality and
integrity of persistent data, clients specify\nper-file policies to Gu
ardat declaratively, concisely and separately from code. \nGuardat\nen
forces policies by mediating I/O in the storage layer. In contrast to
prior \ntechniques,\nwe protect against accidental or malicious circum
vention of higher software \nlayers.\nWe present the design and protot
ype implementation, and demonstrate that Guardat\nefficiently enforces
example policies in a web server.\n \nTo protect the confidentiality
and integrity of in-memory data, ERIM isolates\nsensitive data using I
ntel Memory Protection Keys (MPK), a recent x86 extension\nto partitio
n the address space. However, MPK does not protect against malicious\n
attacks by itself. We prevent malicious attacks by combining MPK with
call gates\nto trusted entry points and ahead-of-time binary inspectio
n. In contrast to \nexisting\ntechniques, ERIM efficiently protects fr
equently-used session keys of web \nservers,\nan in-memory reference m
onitor’s private state, and managed runtimes from native\nlibraries. T
hese use cases result in high switch rates of the order of 10 5 –10 \n
6 switches/s.\nOur experiments demonstrate less then 1% runtime overhe
ad per 100,000 \nswitches/s,\nthus outperforming existing techniques.
TRANSP:OPAQUE
CLASS:PUBLIC
URL:http://domino.mpi-inf.mpg.de/internet/events.nsf/all/20190205-1730
-029
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Berlin:20190206T113000
DURATION:PT0H20M0S
UID:A1BDC4A6B62C1A3FC125838300477F27
DTSTAMP:20190115T140056
SUMMARY:Kurt Mehlhorn (MPI-INF - D1) talks about "D1 Group Meeting"
LOCATION:Saarbrücken building E1 4, room 333
CATEGORIES:AG1 Group Meeting [AG Audience]
CONTACT:Ben Wiederhake
DESCRIPTION:Bla
TRANSP:OPAQUE
CLASS:PUBLIC
URL:http://domino.mpi-inf.mpg.de/internet/events.nsf/all/20190206-1130
-333
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Berlin:20190206T121500
DURATION:PT1H0M0S
UID:18B2FA30C70E2D3AC12582FF003DB4FA
DTSTAMP:20180905T131400
SUMMARY:Adish Singla (Max Planck Institute for Software Systems) talks
about "Machine Teaching"
LOCATION:Saarbrücken building E1 5, room 002
CATEGORIES:Joint Lecture Series [Public Audience]
CONTACT:Jennifer Müller\, Phone: 2900
DESCRIPTION:Much of the research in machine learning has focused on de
signing algorithms \nfor discovering knowledge from data and learning
a target model. What if there \nis a ``teacher" who knows the target a
lready, and wants to steer the ``learner" \ntowards this target as qui
ckly as possible? For instance, in an educational \nsetting, the teac
her (e.g., a tutoring system) has an educational goal that she \nwants
to communicate to a student via a set of demonstrations and lessons;
in \nadversarial attacks known as training-set poisoning, the teacher
(e.g., a \nhacking algorithm) manipulates the behavior of a machine le
arning system by \nmaliciously modifying the training data. This lectu
re will provide an overview \nof machine teaching and cover the follow
ing three aspects: (i) how machine \nteaching differs from machine lea
rning, (ii) highlight the problem space of \nmachine teaching across d
ifferent dimensions, and (iii) discuss our recent work \non developing
teaching algorithms for human learners.
TRANSP:OPAQUE
CLASS:PUBLIC
URL:http://domino.mpi-inf.mpg.de/internet/events.nsf/all/20190206-1215
-002
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Berlin:20190220T113000
DURATION:PT0H20M0S
UID:B096FC0FA2A117E5C1258347002ECC91
DTSTAMP:20181116T093110
SUMMARY:Kurt Mehlhorn (MPI-INF - D1) talks about "D1 Group Meeting"
LOCATION:Saarbrücken building E1 4, room 333
CATEGORIES:AG1 Group Meeting [AG Audience]
CONTACT:Ben Wiederhake
DESCRIPTION:Bla
TRANSP:OPAQUE
CLASS:PUBLIC
URL:http://domino.mpi-inf.mpg.de/internet/events.nsf/all/20190220-1130
-333
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Berlin:20190306T113000
DURATION:PT0H20M0S
UID:6E06B00737166990C1258347002ED49E
DTSTAMP:20181116T093130
SUMMARY:Kurt Mehlhorn (MPI-INF - D1) talks about "D1 Group Meeting"
LOCATION:Saarbrücken building E1 4, room 333
CATEGORIES:AG1 Group Meeting [AG Audience]
CONTACT:Ben Wiederhake
DESCRIPTION:Bla
TRANSP:OPAQUE
CLASS:PUBLIC
URL:http://domino.mpi-inf.mpg.de/internet/events.nsf/all/20190306-1130
-333
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Berlin:20190320T113000
DURATION:PT0H20M0S
UID:AD83047CAD18EC22C1258347002ED98B
DTSTAMP:20181116T093143
SUMMARY:Kurt Mehlhorn (MPI-INF - D1) talks about "D1 Group Meeting"
LOCATION:Saarbrücken building E1 4, room 333
CATEGORIES:AG1 Group Meeting [AG Audience]
CONTACT:Ben Wiederhake
DESCRIPTION:Bla
TRANSP:OPAQUE
CLASS:PUBLIC
URL:http://domino.mpi-inf.mpg.de/internet/events.nsf/all/20190320-1130
-333
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Europe/Berlin:20190403T113000
DURATION:PT0H20M0S
UID:A3E90324C466A08CC1258347002EC5A7
DTSTAMP:20181116T093052
SUMMARY:Kurt Mehlhorn (MPI-INF - D1) talks about "D1 Group Meeting"
LOCATION:Saarbrücken building E1 4, room 333
CATEGORIES:AG1 Group Meeting [AG Audience]
CONTACT:Ben Wiederhake
DESCRIPTION:Bla
TRANSP:OPAQUE
CLASS:PUBLIC
URL:http://domino.mpi-inf.mpg.de/internet/events.nsf/all/20190403-1130
-333
END:VEVENT
END:VCALENDAR