About  Research  Events  Reports  Alumni  Contact  Home
September 1922, 2017
Workshop Community Detection and Network Reconstruction
SPONSORED BY:
SUMMARY
Recent advances in the study of networks in
the physics, mathematics and computer science communities have provided
novel probabilistic tools to address various networkrelated problems of
practical relevance. Two seemingly different, but actually intimately
related, problems are community detection and network reconstruction.
ORGANIZERS
CONFIRMED SPEAKERS
MONDAY September 18
TUESDAY September 19
THURSDAY September 21
FRIDAY September 22
Sebastien Bubeck New results in distributed optimization and consensus learning Consider a graph where each vertex v is associated to a convex
function f_v. Assume that each vertex is a computing unit with oracle
access to its associated function. We consider the distributed
optimization problem of reaching the minimum of f:=\sum_v f_v. We show
how the expansion properties of the graph and the conditioning of f
interact in the description of the fundamental limits for this problem.
A practical decentralized algorithm will also be presented with
experiments for leastsquares and logistic regression. (The talk assumes
no prior knowledge in convex optimization.) Ton Coolen Tailored random graph ensembles: analysis, generation algorithms, and loops In many disciplines one would like to `tailor' (or deform) random
graphs, i.e. build statistical features into their topologies, so that
they can be used as more realistic models of observed realworld
networks. This can be achieved in a systematic and unbiased manner using Fengnan Gao Statistical Estimation of Preferential Attachment Networks The preferential attachment (PA) network is a popular way of modeling
the social networks, the collaboration networks and etc. The PA network
model is an evolving network where new nodes keep coming in. When a new
node arrives, it establishes only one connection with an existing node.
The random choice on the existing node is via a multinomial distribution
with probability weights based on a preferential function $f$ on the
degrees. $f$ is assumed apriori nondecreasing, which means the nodes
with high degrees are more likely to get new connections, i.e., "the
rich get richer". We proposed an estimator on $f$, that maps the natural
numbers to the positive real line. We show, with techniques from
branching process, our estimator is consistent. If $f$ is affine,
meaning $f(k) = k + \delta$, it is well known that such a model leads to
a powerlaw degree distribution. We proposed a maximum likelihood
estimator for $\delta$ and establish the asymptotic normality and
efficiency on the MLE. If the PA function is sublinear and assumes a
parametric form, then we show the maximum likelihood estimator is also
efficient, despite the difficulty in analyzing the likelihood. We will
also talk about some recent advances revealing the connection between
the empirical estimator and the maximum likelihood estimator. PoLing Loh Optimal rates for community estimation in the weighted stochastic block model Community identification in a network is an important problem in fields such as social science, neuroscience, and genetics. Over the past decade, stochastic block models (SBMs) have emerged as a popular statistical framework for this problem. However, SBMs have an important limitation in that they are suited only for networks with unweighted edges; in various scientific applications, disregarding the edge weights may result in a loss of valuable information. We study a weighted generalization of the SBM, in which observations are collected in the form of a weighted adjacency matrix and the weight of each edge is generated independently from an unknown probability density determined by the community membership of its endpoints. We characterize the optimal rate of misclustering error of the weighted SBM in terms of the Renyi divergence of order 1/2 between the weight distributions of withincommunity and betweencommunity edges, substantially generalizing existing results for unweighted SBMs. Furthermore, we present a principled, computationally tractable algorithm based on discretization that achieves the optimal error rate without assuming knowledge of the weight densities.
PRACTICAL INFORMATION ● VenueEurandom, Mathematics and Computer Science Dept, TU Eindhoven, Den Dolech 2, 5612 AZ EINDHOVEN, The Netherlands
Eurandom is
located on the campus of Eindhoven
University of
Technology,
in the
Metaforum building
(4th floor) (about
the building). The university is
located at 10 minutes walking distance from Eindhoven main railway
station (take
the exit north side and walk towards the tall building on the right
with the
sign TU/e).
● RegistrationRegistration is free, but compulsory for speakers and participants. Registration is now open. Please go to: REGISTRATION
● AccommodationFor invited speakers and organizers we will take care of accommodation. Other attendees will have to make their own arrangements. For hotels around the university, please see: Hotels (please note: prices listed are "best available"). Reimbursement available up to 80 euro per night. More hotel options can be found on the webpages of the Tourist Information Eindhoven, Postbus 7, 5600 AA Eindhoven.
● TravelFor those arriving by plane, there is a convenient direct train connection between Amsterdam Schiphol airport and Eindhoven. This trip will take about one and a half hour. For more detailed information, please consult the NS travel information pages. Many low cost carriers also fly to Eindhoven Airport. There is a bus connection to the Eindhoven central railway station from the airport. (Bus route number 401) For details on departure times consult http://www.9292ov.nl The University can be reached easily by car from the highways leading to Eindhoven (for details, see our route descriptions or consult our map with highway connections.
● Conference facilities : Conference room, Metaforum Building MF11&12The meetingroom is equipped with a data projector, an overhead projector, a projection screen and a blackboard. Please note that speakers and participants making an oral presentation are kindly requested to bring their own laptop or their presentation on a memory stick.
● Conference SecretariatUpon arrival, participants should register with the workshop officer, and collect their name badges. The workshop officer will be present for the duration of the conference, taking care of the administrative aspects and the daytoday running of the conference: registration, issuing certificates and receipts, etc.
● CancellationShould you need to cancel your participation, please contact Patty Koorn, the Workshop Officer.
● ContactMrs. Patty Koorn, Workshop Officer, Eurandom/TU Eindhoven, koorn"at"eurandom.tue.nl
Last updated
170717,

