Spring IFAB 2009

Bookmark as: http://www.his.se/infofusion/ifab/spring09

Final program (printer friendly)

Thursday 5 March: 9.30 – 12.00 in The Chancellery (Kanslihuset)

(Presentations and notes are available on IFAB's Internal pages)

09.30 - 10.00  Coffee/tea

10.00 - 12.00  IFAB meeting with IFEC + HS mgmt

Leif Larsson, President (HIS): Welcome address
Sten F Andler: Highlights of 2008 and Strategic business plan
Scenario managers: Scenario highlights of 2008 and Business opportunities

12.00 - 13.15  IFAB lunch with IFEC + HIS mgmt (“Hyllan”)

Thursday 5 March: 13.15 – 17.00 in G211 (Open to all)

Individual project presentations

13.15 - 14.00  Martin Holmberg, Swedish Defence Research Agency (FOI) & Linköping University: Forum Securitatis, FOCUS, and EU FP7 – Examples of FOI commitments in security and crisis management

The graduate school for security and crisis management -- "Forum Securitatis" -- is a co-operation between FOI and the University of Linköping. The first phase runs for four years at around SEK 5 million per year, of which two thirds are funded by Vinnova. The graduate school develops a completely new line of education for security and crisis management, based on an interdisciplinary perspective and developed in co-operation with industries, authorities, and the research centre FOCUS. The FOCUS Institute Excellence Center is active within the area of advanced sensor and sensor network Technology, and is a strategic research initiative hosted by FOI. The FOCUS partnership includes academic as well as industry members. In this talk, a description of Forum Securitatis and FOCUS will be given together with a short glimpse of FOI initiatives on a European level in the area of security and crisis management.

14.00 - 14.30  Martin Smedberg, Saab Microwave Systems: Intelligent surveillance in applications related to civil security

Today's surveillance systems allow us to collect an excessive amount of information from a network of sensors, 24 hours a day and seven days a week. In an intelligent surveillance systems the sensors are linked to computers to detect and trace objects and to analyse unusual behaviour in order to identify threats. In this work we will report from different efforts concerning intelligent surveillance related to civil security performeed at Saab. Special attention will be given to an intelligent video surveillance system and a set of trials performed at Landvetter airport.

14.30 - 15.00  Sandor Ujvari: Information Fusion – The EuroMaint Case

In this presentation an agent based simulation based on EuroMaint case is presented and the result show clear benefits to this type of approach in improved basis for decision-making. This agent  based simulation approach is related to Information fusion  in the way that information from different sources are collected and fused in a synergistic manner into a representation that provides effective support for human decision making (agent based information fusion). The problem of maintenance planning is that operations need to be well balanced in terms of utilization rate of resources, and the balance between maintenance cost and the up-time of the customers' production systems. An MSP also serves more customers with long-term contracts for preventive maintenance and/or require very rapid acute maintenance assistance. To realize efficient and effective maintenance planning in this environment, managers, or decision makers, continuously need correct and updated information from different sources. Moreover, they need to be able to predict the outcome of their decisions and how their decision affects operations.

15.00 - 15.30  Break

15.30 - 16.00  Gunnar Mathiason: Wireless Sensor Network (WSN) testbed

Wireless Sensor Networks (WSNs) play an increasingly more important role in surveillance and situation awareness systems. WSNs monitor parts of the environment, and need to provide timely alarms for any anomalies detected. The Common Goals and Infrastructure project within the Information Fusion (IF) research program at University of Skövde, Sweden, aims at finding generic properties of infrastructures to be used for IF. To enable experiments with new IF algorithms and approaches, we are building up a WSN testbed environment. Using this testbed, we evaluate new IF approaches, in collaboration with our supporting industrial partners. Applications includes military vehicle recognition sensors (Exensor), surveillance situation awareness (SAAB), and soil monitoring for real-time precision agriculture (SLU/Agroväst). These companies already support the Information Fusion Program as external sponsors, but also want to deepen the collaboration regarding related applications and infrastructure issues. We also collaborate with University of Latvia with a group that specialize in sensor integration and software/hardware development for sensor platforms. Further, we have collaboration with the WSN group at University of Virginia. This group is one of the top three groups in the US for state-of-the art research in WSNs. The sensor data is made available on small mobile tablet PCs, in a distributed database, and sensor data is propagated between a set of database nodes. By demonstrating the capabilities of such a distributed information fusion setup, approaches and feasibility can be evaluated. A demonstrator also inspires further collaboration and research in distributed fusion.

16.00 - 16.30  Maria Nilson: A method of capturing the interaction between human and technology in the context of information fusion

Within information fusion there is an implicit drive towards increasing automation. It is also acknowledged that research on the user side of information fusion has been relatively neglected. In this context, the presented research aims at investigating user issues within a information fusion setting, and thereby, enhance the knowledge of such issues. In particular, the aim is to explore how current/existing interaction between humans and technology can be captured in order to enable future increase of automation (in the studied work environment). Increasing the automation of  a manual process is not straightforward;  in addition to fusing the data captured in the physical and digital artefacts (e.g., information from a fax or database, a piece of paper with a list of attributes, radar readings etc.), the artefacts itself serves as cognitive support (e.g., external memory) which need to be accounted for. This aspect is not typical captured by current methods, hence, the presented research will develop a method based on the theory of distributed cognition to be able to account for such implicit issues. The proposed method includes a theoretical grounding in cognitive science, a step-to-step instruction, and a notation, to be used by practitioners for capturing interaction within information fusion processes.

16.30 - 17.00  Joeri van Laere: Maritime Domain Awareness (MDA) workshop – A comparison of taxonomies for maritime surveillance anomalies

During autumn 2008 the University of Skövde, Saab and the Swedish Defense Research Agency FOI conducted a pre-study on World Wide Maritime Surveillance. Part of this study was a one day workshop with stakeholders to identify potential anomalies in WWMS during a computer supported brainstorming session. This talk discusses an evaluation of this workshop. First, the output of the workshop is compared with the output of a similar workshop as presented in Roy (2008, SPIE paper 6945-28). Secondly, methodological differences are discussed between capturing know-how of subject matter experts in a brainstorming session, versus observing daily work of operators, as applied in Nilsson (2008, Fusion conference). Preliminary conclusions are that the outcomes of both knowledge capturing strategies heavily depend on the selection of operators to be observed, as well as the selection of subject matter experts invited. Therefore, multiple studies and workshops are needed until no ‘new’ anomalies are identified.

18.00 -  Dinner for IFAB and IFEC members

Friday 6 March: 08.45 – 10.45 in G211 (Open to all)

08.45 - 09.15  Joeri van Laere: Challenges for IF performance evaluation

Good performance implies good information fusion that enables good decision making. But what is ‘good’ may differ from situation to situation. Sometimes it may be more important to be fast than to be accurate, sometimes it may be the other way around. Even more problematic is that the target value of a single performance indicator may fluctuate due to changing decision situation characteristics.  What was fast, valid or accurate for a certain decision yesterday or a year ago may be slow, irrelevant or imprecise for the same decision today. Finally, it may be impossible to define ‘good’ performance at all. In practical applications, where we often do not know ground truth, it may be difficult to find reasonable target values for performance indicators.
This talk illustrates which performance evaluation problems can occur in practice by discussing a hypothetical example. Secondly, results are presented from an analysis of about 40 articles from the information fusion journal and the last two Fusion conferences, to show how current research on fusion performance evaluation addresses such performance evaluation problems. Finally, some recommendations are developed, arguing for more flexibility and robustness in performance measurement, rather than accurateness.

09.15 - 09.45  Tuve Löfström: Utilizing Diversity and Performance Measures for Ensemble Creation

An ensemble is a composite model, aggregating multiple base models into one predictive model. An ensemble prediction, consequently, is a function of all included base models. Both theory and a wealth of empirical studies have established that ensembles are generally more accurate than single predictive models. The main motivation for using ensembles is the fact that combining several models will eliminate uncorrelated base classifier errors. This reasoning, however, requires the base classifiers to commit their errors on different instances – clearly there is no point in combining identical models. Informally, the key term diversity means that the base classifiers commit their errors independently of each other. The problem addressed in this thesis is how to maximize ensemble performance by analyzing how diversity can be utilized when creating ensembles. A series of studies, addressing different facets of the question, is presented. The results show that ensemble accuracy and the diversity measure difficulty are the two individually best measures to use as optimization criterion when selecting ensemble members. However, the results further suggest that combinations of several measures are most often better as optimization criteria than single measures. A novel method to find a useful combination of measures is proposed in the end. Furthermore, the results show that it is very difficult to estimate predictive performance on unseen data based on results achieved with available data. Finally, it is also shown that implicit diversity achieved by varied ANN architecture or by using resampling of features is beneficial for ensemble performance.

09.45 - 10.15  Break

Friday 6 March: 10.15 – 12.45 in The Chancellery (Kanslihuset)

10.15 - 11.30  IFAB private reflection  ||  IFEC work on business plan (in parallel in IFVL)

11.30 - 12.45  IFAB feedback meeting with IFEC + HS mgmt

13.00 - 14.00  Lunch (Hyllan) or 13.00 - 15.00 Portalen (invited guests only) (*)

(*) taklagsfest = topping out ceremony (of new research building, Portalen = The Portal) 

IFAB = Infofusion Advisory Board
IFEC = Infofusion Executive Committee

Uppdaterad: 2011-05-03
Sidansvarig: Marcus Brohede