Programme
Outline of the programme
All events will take place in Lecture hall F211 (address: Unioninkatu 38), unless otherwise stated.
Sunday, August 7
Monday, August 8
- 9:00-17:00: Technical sessions (F211)
- 18:00-20:00: Reception offered by the Rector of the University of Helsinki
Location: University Main Building, (Room "Lehtisali"),
Unioninkatu 34
Tuesday, August 9
- 9:00-17:00: Technical sessions (F211)
- 18:15: Sauna and Banquet Dinner at
Restaurant
Uunisaari.
Location: Uunisaari Island, (address:
Merisatamanranta/Kompassitori)
Wednesday, August 10
- 9:00-12:00: Technical sessions (F211)
Technical Programme
Sunday, August 7
16:00-16:30 Opening of the Workshop
16:30-17:30 Plenary Talk: Mati Wax
17:30-20:00 Cocktails
Monday, August 8
9:00-10:00 Plenary Talk: Neri Merhav
10:00-10:25 Coffee Break
10:25-12:05 Session 1
- Vincent Poor: Adaptive Sampling for Sparse Recovery
- Andrew Barron: Sparse Superposition Codes of Low Complexity and
Exponentially Small Error Probability at all Rates below Capacity for
the Gaussian Channel
- Garvesh Raskutti: Minimax Optimal Rates for Sparse Additive Models for
Kernel Classes
- Marcelo Weinberger: Deinterleaving Markov Processes via Penalized
Maximum-Likelihood
12:05-14:05 Lunch
14:05-15:20 Session 2
15:20-15:45 Coffee Break
15:45-17:00 Session 3
18:00-20:00 Rector's Reception
Tuesday, August 9
09:00-10:00 Plenary Talk: Veronica Gonzalez-Lopez and Jesus Garcia
10:00-10:25 Coffee Break
10:25-12:05 Session 4
12:05-14:05 Lunch Break
14:05-15:20 Session 5
15:20-15:45 Coffee Break
15:45-17:00 Session 6
18:15-22:00 Sauna and Banquet Dinner
Wednesday, August 10
09:00-10:20 Session 7
10:20-10:45 Coffee Break
10:45-12:00 Session 8
12:00-14:00 Lunch
Plenary speakers
Mati Wax (Wavion Wireless Networks):
Position Location by Multipath Fingerprinting
|
- Abstract: Position location of a wireless source is a well known
and old problem, with both military and commercial applications. Many
techniques have been proposed and applied to this problem over the
last 70 years. All these techniques are based on the assumption that
the wireless signal travels from the source to the receiving antennas
along the line-of-sight path connecting them. Unfortunately, in urban
canyons and in indoors venues such as warehouses, hospitals,
production floors, malls, etc., this is usually never the case. In
these cases the propagation from the wireless source to the receiving
antennas is usually made through reflections from buildings and walls,
referred to as multipath, which may be very different from the
line-of-sight path. As a result, the applicability of the classical
position location techniques to such scenarios is void.
This talk presents a new method for position location, specifically
tailored to heavy multipath scenarios, which exploits the multipath to
its advantage, rather than suffers from it. The basic underlying
principle of this method is that in heavy multipath scenarios there is
a one-to-one correspondence between the multipath "fingerprint", as
captured by an array of receiving antennas, and the location of the
source. Based on this premise, the position location problem is casted
as a pattern matching problem: the multipath "fingerprint" of the
source to be located is extracted, compared to a data base of
"fingerprints", which are pre-collected from the target area, and the
best match is selected as the source location. We will present various
theoretical, computational and practical aspects of the method and
discuss its pros and cons.
- Time: 16:30-17:30, Sunday, August 7
- Location: Lecture hall F211, Unioninkatu 38
- Abstract: Following a brief introduction of
basic background in statistical physics, we will describe
relationships and analogies between certain models of spin glasses, in
particular - the random energy model (REM), and the behavior of
certain ensembles of codes for communication systems. Beyond the
purely theoretical aspects of these relations, we will also
demonstrate how analysis techniques, rooted in the statistical
mechanics of the REM, can be harnessed to obtain sharper and more
accurate evaluations of the ensemble performance of these codes. Time
permits, we will also point out several extensions of the basic model.
- Time: 9:00-10:00, Monday, August 8
- Location: Lecture hall F211, Unioninkatu 38
- Abstract: In this work we introduce a new and richer class of
finite order Markov chain models and address the following model
selection problem, find the Markov model with the minimal set of
parameters (called here minimal Markov model) which is enough to
represent a source as a Markov chain of finite order. Let us call M
the order of the chain and A the finite alphabet, to determine the
minimal Markov model, we define an equivalence relation on the state
space AM, such that all the sequences of size M
with the same
transition probabilities are put in the same part. In this way we have
one set of (|A|-1) transition probabilities for each part, obtaining
a model with a minimal number of parameters. We show that the model
can be selected consistently using the Bayesian information
criterion. The application of our model is exemplified in two DNA data
sets and on simulated data.
- Time: 9:00-10:00, Tuesday, August 9
- Location: Lecture hall F211, Unioninkatu 38
Neri Merhav
(Technion): Random Coding and Statistical Physics