Technically Co-Sponsored by
EMI        
                                                                                                                                            

Keynote Speakers and tutorials

Keynotes : Dr. Niall Adams, Pr. Malik Ghallab                Tutorial : Dr. Leandro Minku

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Speaker Dr Niall Adams
Title

Big Data in Cyber-Security: Host-Based IP Flow Monitoring using Adaptive Estimation

Dr Niall Adams has been a faculty member of the Department of Mathematics, Imperial College London since 2001, being promoted to the position of reader in 2009. In 2010, he was recipient of a Winton Capital Research prize.  At present, Adams is on extended secondment at the Heilbronn Institute for Mathematical Research, University of Bristol, leading a small team of statistics and data mining researchers. His research interests include classification and data mining, in diverse applications including cyber-security, consumer finance, and nuclear biology. Adams has published numerous articles in peer-reviewed journals and conferences, and  is co-editor of a forthcoming  book entitled "Data Analysis for Network Cyber-Security".

Abstract

To begin, we review data analysis  challenges in network cyber-security, particularly in relation to NETFLOW data. The scale and stochastic structure of such data raises both computational and statistical difficulties. We develop change detectors, that operate at a host level, based on adaptive estimation procedures. The performance of these detectors is assessed with extensive simulation studies, and demonstrated on NETFLOW data collected at Imperial College.

 
Speaker Pr. Malik Ghallab
Title

Models and Algorithms for Deliberate Action in Robotics

Dr. Malik Ghallab is a senior research scientist at LAAS-CNRS and University of Toulouse (Directeur de recherche). He is a member of Hassan II Academy of Science and Technology of Morocco. His research activity is mainly focused on planning, acting and learning in robotics and AI. He contributed to topics such as object recognition and pattern matching, scene interpretation, heuristics search, unification algorithms, knowledge compiling, temporal reasoning, planning, monitoring, and machine learning of robots skills and models of behaviors.

Malik Ghallab was head of the French national interdisciplinary research program in Robotics (2000-2006), director of LAAS-CNRS (2003-2007) and Executive Officer for Research and Technology of INRIA (2007- 2010)

Abstract

Robots are built with inaccurate and noisy sensors and actuators, as well as with partial and uncertain models. Even a low-level reactive command requires prediction capabilities for a robust execution. A robot working in a fixed, well-modeled environment (e.g., robots in manufacturing applications), or a robot designed for a single task (e.g., vacuum cleaning or lawn-mowing robot) will have to implement these low-level predictive commands. An autonomous robot facing a variety of tasks and environments requires significantly more extended planning and deliberation capabilities.

A deliberate action is a purposeful, planned activity, pursued and carried out in order to achieve some objectives. Several computational functions are needed to endow a robot with deliberation capabilities, among which: planning, refining planned task into commands and controlling their execution, observing the context, monitoring discrepancies between prediction and observation, assessing the relevance of current objectives and learning how to improve its behavior. These functions implement different mathematical models and algorithms to perform anticipation, decision and control. They rely on a wide spectrum of representations, running from differential geometry and dynamics, to Temporal Action Logic, Dynamic Bayes Networks, or Temporal Markov Decision Processes. This talk will illustrate some of these representations, models and algorithms used for addressing problems such as, for example:

Prediction in dynamic control; Prediction in motion planning; Prediction in task planning; Prediction in observing and monitoring; Prediction in assessing situations and recognizing plans.

It will briefly discuss how corresponding deliberation functions are related and how they can be organized within an autonomous robot.

 

 
Speaker

Dr. Leandro Minku

Title

Online learning in changing environments

 Leandro L. Minku is a Research Fellow at the Centre of Excellence for Research in Computational Intelligence and Applications (CERCIA), School of Computer Science, the University of Birmingham (UK). He received the BSc, MSc and PhD degrees in Computer Science from the Federal University of Parana (Brazil) in 2003, the Federal University of Pernambuco (Brazil) in 2006, and the University of Birmingham (UK) in 2011, respectively. He was an intern at Google Zurich for six months in 2009/2010, the recipient of the Overseas Research Students Award (ORSAS) from the British government, and of several scholarships from the Brazilian Council for Scientific and Technological Development (CNPq).

Dr. Minku's main research interests are machine learning in changing environments, ensembles of learning machines and software prediction models. His work has been published in internationally renowned journals such as IEEE Transactions on Knowledge and Data Engineering, ACM Transactions on Software Engineering and Methodology, and Neural Networks.

Abstract

Most machine learning algorithms operate in offline mode. They first learn how to perform a certain task, and then are used to perform this task. However, most practical problems change with time.

For example, the problem of predicting users' preferences in information filtering systems may involve changes in users' preferences; the problem of classifying webpages may involve changes in the most representative words of different webpage categories; the problem of credit card approval may involve changes in customers' reliability. Different from offline learning algorithms, online learning algorithms can be used to adapt to such changes based on newly incoming training examples.

These algorithms do not have a separate training and testing phase, but learn throughout their lifetime as they are used to perform a certain task. Due to the practical need for adaptive learning systems, there has been an increasing number of works on online learning algorithms able to operate in changing environments. This talk will first introduce some basic concepts in this field. Then, different approaches for online learning in changing environments will be described, in particular online ensemble learning approaches able to deal with different types of change, and novel approaches able to operate in changing environments with class imbalance. Finally, some future research directions will be discussed

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 


EMI     FSTM ENSIAS INPT
UM5A UH2M UM5S ANRT
     
homepage | contact | 2012 SITA