Agenda

Conferences

WIC Midwinter Meeting on Deep Learning

Organized by Werkgemeenschap voor Informatie- en Communicatietheorie, and IEEE Benelux Chapter on Information Theory

Additional information ...


MSc Thesis Presentation

Blind Graph Topology Change Detection: A Graph Signal Processing Approach

Ashvant Mahabir

Graphs are used to model irregular data structures and serve as models to represent/capture the interrelationships between data. The data in graphs are also referred as graph signals. Graph signal processing (GSP) can then be applied which basically extends classical signal processing to solve problems. Anomaly detection is an example of such a problem. Two hypothetical situations are given, and a detector has to be designed to distinguish between these. Under the null hypothesis, graph structures are considered to be untouched. Under the alternative hypothesis, (unknown) topological changes might have occurred. Now by incorporating a priori knowledge about the graphs, the decision making process should improve. In most works, a priori knowledge of the graphs under the null and alternative hypothesis was incorpo- rated. This means that detectors were designed which were able to anticipate on possible topological changes. In this thesis, the problem is considered where only a priori knowledge of the graph under the null hypothesis is exploited. This means that detectors are not able to anticipate on potential changes and this where blind detection comes into play. Blind detection is important because it considers a more realistic scenario. In this work, the blind topology change detector (BTCD) and the constrained blind topology change detector (CTCD) are derived which exploit different properties of the data re- lated to the known graph structure. For the BTCD, the bandlimitedness of graph signals was exploited and for the CTCD, the graph signal smoothness. The main question in this work, was to find out what the potentials are with the blind detection principle for graph change detection. Different test scenarios are used to evaluate the detectors on both synthetic and real data. For the BTCD, the obtained results compare well when information about the alternative graph is available. For this detector, the potential of blind detection was highly visible. For bandlimited graph signals, the BTCD as good as detectors using full information. For the CTCD, comparable results (with detectors using full information) are attained for just a few test scenarios. For small changes, the graph signal smoothness seems to be less powerful as to the graph signal bandlimitedness. This study showed that graph change detection is still possible without having full information. Some graph signal properties are more powerful w.r.t. others.


Signal Processing Seminar

When is Network Lasso Accurate: The Vector Case

Nguyen Tran

A recently proposed learning algorithm for massive network-structured data sets (big data over networks) is the network Lasso (nLasso), which extends the well- known Lasso estimator from sparse models to network-structured datasets. Efficient implementations of the nLasso have been presented using modern convex optimization methods. In this paper, we provide sufficient conditions on the network structure and available label information such that nLasso accurately learns a vector-valued graph signal (representing label information) from the information provided by the labels of a few data points.


Signal Processing Seminar

Semi-supervised learning for likelihood-based classifiers

Marco Loog
Bioinformatis/Pattern Recognition group

Bio: Marco Loog received an M.Sc. degree in mathematics from Utrecht University and in 2004 a Ph.D. degree from the Image Sciences Institute for the development and improvement of contextual statistical pattern recognition methods and their use in the processing and analysis of images. After this joyful event, he moved to Copenhagen where he acted as assistant and, eventually, associate professor next to which he worked as a research scientist at Nordic Bioscience. In 2008, after several splendid years in Denmark, Marco moved to Delft University of Technology where he now works as an assistant professor in the Pattern Recognition Laboratory. He currently is associate editor of Pattern Recognition and honorary full professor in pattern recognition at the University of Copenhagen. Marco's research interests primarily include all types of variations to supervised learning.


Signal Processing Seminar

Wangyang Yu

Additional information ...


Signal Processing Seminar

The Quest for Fast Learning from Few Examples

Andreas Loukas

Though the data in our disposal are numerous and diverse, deriving meaning from them is often non trivial. This talk centers on two key challenges of data analysis, relating to the sample complexity (how many examples suffice to learn something with statistical significance) and computational complexity (how long does the computation take) of learning algorithms. In particular, we are going to consider two famous unsupervised algorithms, principal component analysis and spectral clustering, and ask what can they learn when given very few examples or a fraction of the computation time.


Signal Processing Seminar

Mario Coutiño Minguez