mlsplogo MLSP2015
IEEE International Workshop on
Machine Learning for Signal Processing

September 17-20, 2015  Boston, USA

Special Sessions

Information-Centric Machine Learning for Signal Processing
The booming of internet and mobile communications opens a new era for machine learning and signal processing that aims to efficiently acquire useful information from big networked data. In particular, an emerging trend in signal processing and machine learning advocates the information-centric analyzations of data analytics and knowledge discovering of the data. Comparing to traditional approaches, the information-centric signal processing and machine learning focus on studying the fundamental aspects of the information fusion, inference, and prediction of networked data, from communication and information theoretic prospectives. By leveraging the concept of information exchanging and the statistical frameworks from the communication and information theories, the information-centric signal processing aims to identify the information flows and the knowledge structure of the collected data, and the acquired information structure is further applied to enhance the efficiency and performance of the filter designs in signal processing tasks. In addition, the information-centric signal processing results are recently employed to investigate machine learning problems, such as the data clustering, pattern recognition, feature selection, and community detection. This leads to promising research topics that combine scientific and engineering subjects in communication, signal processing, and machine learning, and the corresponding studies also suggest practical algorithm designs for efficiently implementing knowledge structure discovery and feature selection of the data. In a nutshell, the information-centric signal processing involves the interplay between communication/information theory, signal processing, and machine learning, and the challenging research problems have gained sufficient attentions from our research community.

Scope and Topics of Interest

The proposed special session covers the technical aspects of the interplay between information/communication theory, signal processing, and machine learning, and more specifically will focus on original contributions regarding the applications of the statistical framework of information and communication theory to machine learning and data analytics. It aspires to raise holistically the awareness of the research and industrial communities in the areas of both information theory and machine learning from diverse and broad perspectives and provide a venue for the presentation and dissemination of the latest original contributions in the corresponding field of science. The aim of this special session is to provide a forum that brings together scientists and researchers from all over the world to present their cutting-edge innovations in all aspects of the related research fields. The main topics include but are not limited to:
  • Information and communication theoretic data analytics
  • Information-centric data fusion, inference, and prediction
  • Privacy preserving data inference
  • Stochastic models of community detections
  • Kernel methods with information theoretic approaches
  • Other applications of information theory and communication theory to machine learning and signal processing

Organized by Shao-Lun Huang, National Taiwan University and Massachusetts Institute of Technology,

Machine Learning and Signal Processing for Cybersecurity Analytics
“Secure Cyberspace” is one of the National Academy of Engineering Grand Challenges for Engineering. There is a lot of attention on this as the media continues to present headline after headline regarding breeches and cyber risks. Comparing to other domains, the challenges in cybersecurity analytics include large-scale data, real-time streaming, evolving threat patterns, highly-skewed data and privacy preservation. In addition to traditional learning algorithms, graphical modeling and deep learning have been applied in this domain. There are a number of academic and industrial organizations applying machine learning, signal processing, pervasive computing towards this problem space – and the findings are very promising. The topics cover threat intelligence, access control, privacy preservation and data protection. The capabilities within this area reach across a wide range of very solid niche solutions that allow much to be learned in the research as to how to apply Machine Learning and Signal Processing (MLSP) techniques to solve particular problems associated with this grand challenge. We feel that it is very timely to bring the research together in a special session in order to further findings and explore new and innovative approaches toward securing cyberspace.

Session scope and main topics

The session scope will be open to include all aspects of Cybersecurity Analytics. We would like to have presentations provided that focus on multiple aspects of MLSP – including the application of different machine learning and signal processing algorithms. We would like the papers and presentations to drill in on the techniques used within this problem space. We welcome applications across all aspects of Cybersecurity – from log monitoring to real time network behavioral monitoring, from anomaly detection to advanced classification techniques, from access control and identity to privacy preservation and data protection. The main topics include but are not limited to:
  • Learning techniques for log monitoring and threat classification
  • Feature selection for anomaly detection
  • Graphical models and big data for real-time network behavioral monitoring
  • Other applications of MLSP for determining baseline and normal in an infected environment, Breech detection, Access control, Privacy preservation and data protection

Organized by Catherine Huang, Intel Labs, and Paula Greve, McAfee,

Powered by CONWIZ, © Copyright 2014. Page maintained by Jan Larsen (Email)