Program

All events will take place in Room t-014, unless otherwise stated.

Wednesday, June 24

13.30-14.00: Registration and opening of the workshop
14.00-15.00: Teemu Roos: A tutorial type introduction to MDL
15.00: Coffee break
15:30-16.00: Flemming Topsøe: Generalization of proper scoring rules
16.00-16.30: Łukasz Dębowski: Hilberg Processes and Nonexistence of Universal Redundancy Ratios
17.30-19.00: Welcoming reception
Location: Rundetårn

Thursday, June 25

8.30-9.30am: Plenary talk: Steffen Lauritzen: Proper local scoring rules
9.30: Coffee break
10.00-10.30: Joseph O'Sullivan: Fundamental Information Geometry Problems. Theory and Applications.
10.30-11.00: Peter Harremoës: Applications of proper score and sufficiency
11.00-11.30: Mikhail Malyutov, P. Grosu and T. Zhang: SCOT modeling, training and Homogeneity testing
11.30-12.00: Boris Ryabko and Nadezhda Savina: The one-time pad is robust to small deviations from randomness.
12:00: Lunch in the main building
13.30-14.00: Peter Beelen: Fast decoding of algebraic codes
14.00-14.30: Søren Forchhammer and Huynh van Luong: Distributed source coding of video
14.30-15.00: Ioan Tabus: Context coding of 3D images
15.00: Coffee break
15.00-16.30: Recent results: We divide in a communication group and a statistics group and make short blackboard presentations of exciting new results.
17.20-22.00: Banquet dinner
Location: Trekroner

Friday, June 26

8.30-9.30: Plenary talk: Gerhard Kramer: Upper Bounds on the Capacity of Fiber Channels
9.30: Coffee break
10.00-10.30: Darko Zibar: Opportunities for machine learning techniques in photonics communication
10.30-11.00: Erik Agrell and Magnus Karlsson: Single-user capacity of optical multiuser channels
11.00-11.30: Metodi Yankov og Søren Forchhammer: Achievable information rates on linear interference channels with discrete input
11.30-12.00: Jørn Justesen: Decoding of product codes
12.00: Lunch in the main building
13:00-13.30: Mokshay Madiman: and Ioannis Kontoyiannis: The Ruzsa divergence on Groups
13.30-14.00: Ioannis Kontoyiannis and Maria Skoularidou: Causality and Directed Information Estimation as a Hypothesis test
14.00-14.30: Bo Markussen: Relevance sampling
14.30: Coffee break
14.30-15.00: Kazuho Watanabe: Rate-distortion analysis for kernel-based distortion measures
15:00-15.30: Closing of the workshop
15:30-17:30: Optional excursion to the optical communication lab at DTU organized by Søren Forchhammer and Michael Galili.

Saturday, June 27
There will be an optional excursion organized by Peter Harremoës. This excursion is not included in the registration fee. We will go to Roskilde where we will first visit the Viking Ship Museum. After that we will visit Roskilde Domkirke that has the burial site for Danish kings and queens since Harald Bluetooth (died around 987 AD, the bluetooth protocol is named after him).

Plenary speakers

Steffen Lauritzen (Univ. Copenhagen): Proper local scoring rules

Abstract: A scoring rule is a loss function measuring the quality of a quoted probability distribution Q for a random variable X, in the light of the realized outcome x of X; it is proper if the expected score, under any distribution P for X, is minimized by quoting Q = P. Using the fact that any differentiable proper scoring rule on a finite sample space X is the gradient of a concave homogeneous function, we consider when such a rule can be local in the sense of depending only on the probabilities quoted for points in a nominated neighborhood of x. Under mild conditions, we characterize such a proper local scoring rule in terms of a collection of homogeneous functions on the cliques of an undirected graph on the space X. We also mention proper scoring rules for continuous distributions on the real line. Here we allow further dependence on a finite number m of derivatives of the density at the outcome, and describe a large class of such m-local proper scoring rules.
Time: 8:30am - 9:30am, Thursday June 25

Gerhard Kramer (Tech. Univ. München): Upper Bounds on the Capacity of Fiber Channels

Abstract: The capacity of optical fiber channels seems difficult to compute or even bound. The best capacity lower bounds are based on numerical simulations using the split-step Fourier method. We review a recent capacity upper bound that applies two basic tools to this method: maximum entropy under a correlation constraint and Shannon’s entropy power inequality (EPI). The main insight is that the non-linearity that is commonly used to model optical fiber propagation does not change the differential entropy of a signal. As a result, the spectral efficiency of fiber is at most log(1+SNR), where SNR is the receiver signal-to-noise ratio. The results extend to other channels, including multi-mode fiber.
Time: 8.30am - 9.30pm, Friday June 26

For Participants

Programme

Submissions

Misc

Program

Plenary speakers