Stability of feature selection algorithms: a study on high-dimensional spaces

With the proliferation of extremely high-dimensional data, feature selection algorithms have become indispensable components of the learning process. Strangely, despite extensive work on the stability of learning algorithms, the stability of feature selection algorithms has been relatively neglected. This study is an attempt to fill that gap by quantifying the sensitivity of feature selection algorithms to variations in the training set. We assess the stability of feature selection algorithms based on the stability of the feature preferences that they express in the form of weights-scores, ranks, or a selected feature subset. We examine a number of measures to quantify the stability of feature preferences and propose an empirical way to estimate them. We perform a series of experiments with several feature selection algorithms on a set of proteomics datasets. The experiments allow us to explore the merits of each stability measure and create stability profiles of the feature selection algorithms. Finally, we show how stability profiles can support the choice of a feature selection algorithm. Alexandros Kalousis received the B.Sc. degree in computer science, in 1994, and the M.Sc. degree in advanced information systems, in 1997, both from the University of Athens, Greece. He received the Ph.D. degree in meta-learning for classification algorithm selection from the University of Geneva, Department of Computer Science, Geneva, in 2002. Since then he is a Senior Researcher in the same university. His research interests include relational learning with kernels and distances, stability of feature selection algorithms, and feature extraction from spectral data. Julien Prados is a Ph.D. student at the University of Geneva, Switzerland. In 1999 and 2001, he received the B.Sc. and M.Sc. degrees in computer science from the University Joseph Fourier (Grenoble, France). After a year of work in industry, he joined the Geneva Artificial Intelligence Laboratory, where he is working on bioinformatics and datamining tools for mass spectrometry data analysis. Melanie Hilario has a Ph.D. in computer science from the University of Paris VI and currently works at the University of Geneva’s Artificial Intelligence Laboratory. She has initiated and participated in several European research projects on neuro-symbolic integration, meta-learning, and biological text mining. She has served on the program committees of many conferences and workshops in machine learning, data mining, and artificial intelligence. She is currently an Associate Editor of theInternational Journal on Artificial Intelligence Toolsand a member of the Editorial Board of theIntelligent Data Analysis journal.     

Modeling adaptive behaviors in Context UNITY

Context-aware computing refers to a paradigm in which applications sense aspects of the environment and use this information to adjust their behavior in response to changing circumstances. In this paper, we present a formal model and notation (Context UNITY) for expressing quintessential aspects of context-aware computations; existential quantification, for instance, proves to be highly effective in capturing the notion of discovery in open systems. Furthermore, Context UNITY treats context in a manner that is relative to the specific needs of an individual application and promotes an approach to context maintenance that is transparent to the application. In this paper, we construct the model from first principles, introduce its proof logic, and demonstrate how the model can be used as an effective abstraction tool for context-aware applications and middleware.     

Inserting virtual pedestrians into pedestrian groups video with behavior consistency

In this paper, we propose a novel approach to integrate virtual pedestrians into a scene of real pedestrian groups with behavior consistency, and this is achieved by dynamic path planning of virtual pedestrians. Rather than accounting for the local collision avoidance only, our approach is capable of finding an optimized path for each virtual pedestrian on his way based on the current global distribution of the real groups in the scene. The big challenge is due to the information of both position and velocity of real pedestrians in the video being unavailable; also the distribution of the groups in the scene may vary dynamically. We therefore need to detect and track real pedestrians on each frame of the video to acquire their distribution and motion information. We save this information by an efficient data structure, called environment grid. During the way of a virtual pedestrian, the respective agent frequently emits the detection rays through the environment cells to find the situation of the real pedestrians ahead of him and adjust the original path if necessary. Virtual pedestrians are merged into the video finally with the occlusion between virtual characters and the real pedestrians correctly presented. Experiment results on several scenarios demonstrate the effectiveness of the proposed approach.     

Using Task-Induced Pupil Diameter and Blink Rate to Infer Cognitive Load

Minimizing user cognitive load is suggested as an integral part of human-centered design, where a more intuitive, easy to learn, and adaptive interface is desired. In this context, it is difficult to develop optimal strategies to improve the design without first knowing how user cognitive load fluctuates during interaction. In this study, we investigate how cognitive load measurement is affected by different task types from the perspective of the load theory of attention, using pupil diameter and blink measures. We induced five levels of cognitive load during low and high perceptual load tasks and found that although pupil diameter showed significant effects on cognitive load when the perceptual load was low, neither blink rate nor pupil diameter showed significant effects on cognitive load when the perceptual load was high. The results indicate that pupil diameter can index cognitive load only in the situation of low perceptual load and are the first to provide empirical support for the cognitive control aspect of the load theory of attention, in the context of cognitive load measurement. Meanwhile, blink is a better indicator of perceptual load than cognitive load. This study also implies that perceptual load should be considered in cognitive load measurement using pupil diameter and blink measures. Automatic detection of the type and level of load in this manner helps pave the way for better reasoning about user internal processes for human-centered interface design.     

Sufficient LMI stability conditions for Lur’e type systems governed by a control law designed on their Euler approximate model

In this paper, the stability issue of Lur’e systems governed by a control law stabilising their forward Euler approximate model is investigated. More specifically, the considered control law is obtained by exploiting the advantages of a new Lur’e type Lyapunov function with disconnected level sets. This Lyapunov function is adapted to discrete-time Lur’e systems and to the structure of the forward Euler approximate model. The main result consists of linear matrix inequality conditions allowing to guarantee that the continuous-time Lur’e system associated with the proposed digital control law is globally asymptotically stable. The relevance of this approach is illustrated using a numerical example.     

虚拟群体与动态视频场景的在线实时融合

将虚拟人群与实拍视频场景融合实现虚拟人群对现实场景中动态目标的合理避让,是增强现实技术有待解决的一个新问题.文中首先对实拍视频流进行分析,获取真实动态目标在三维空间中的运动信息;然后将动态目标的运动转化为每个虚拟角色的约束条件,实时更新虚拟角色的运动决策,实现其对现实场景中动态目标的避让以及虚拟角色之间的避让.为了清晰地展示现实世界中多个目标的空间位置关系,还对视频画面中的目标体做了精确的实时分割,以产生正确的遮挡关系和目标定位.最后通过实验结果表明了该算法的有效性.     

Case Study: Retrofitting Large Bridge Piers on the Nakdong River, South Korea

The Gupo Bridge crosses the Nakdong River near the city of Busan, South Korea. During Typhoon Maemi in 2003, the old Gupo Bridge collapsed due to excessive pier scour. More recently, the highway construction on the left-bank floodplain required right-bank channel widening to restore the channel flood-carrying capacity. This 7?m deep floodplain excavation is expected to cause significant local scour around the 8–10?m wide and 3?m thick spread footings of Piers 11 and 12 of the Subway Bridge and Piers 15 and 16 of the Gupo Bridge. Three design options are examined for retrofitting floodplain bridge piers with massive spread footings. A solution with sheet piles and riprap was recommended in 2006 as the most appropriate design, but Plan III with a conical riprap structure around the footings was ultimately constructed in 2007 for economic reasons. Laboratory experiments also highlight the need to place gravel and synthetic filters under the designed riprap.     

Real-space tight-binding approach to electronic structure and stability in substitutional alloys

A real-space approach based on the tight-binding approximation is proposed for studying electronic structure properties, stability and order in substitutional multi-component chemically random alloy based on periodic as well as topological disordered lattices. We show that the coherent potential approximation (CPA) equations including Shiba's off-diagonal disorder can be solved self-consistently in real-space with the same accuracy currently achieved in reciprocal space. An effective one-electron Green function is given by a continued fraction expansion, and the associated effective medium is used to determine the effective cluster interactions which enter the expression of the configurationl part of the total energy for describing order-disorder phenomena in alloys. Some applications will be presented.     

Influence of moisture content on the specific methanogenic activity of dry mesophilic municipal solid waste digestate

The objective of this study was to evaluate the influence of moisture content on the specific methanogenic activity (SMA) of a fresh dry mesophilic digestate from a municipal solid waste digester plant. For this purpose, SMA tests were performed under mesophilic conditions in 500 mL glass bottles of volume used as batch reactors, during a period of 20–25 days. Cellulose, propionate and acetate were used as substrates (5 g_COD kg^?1 digestate) at four different moisture contents, ranging from 65 to 82%. The moisture content strongly influenced the specific methanogenic activity. The highest SMA values were observed at a moisture content of 82% (11.1, 7.8 and 6.0 mg_COD g_VS^?1 d^?1 for cellulose, propionate and acetate spikes, respectively). SMA and moisture content were found to be linearly linked. Dry digestion at low water content is thus detrimental to the biological activity, probably due to physical limitations. Copyright © 2011 Society of Chemical Industry