Data-driven clustering : new methods and applications

Abstract : This work is concerned with the development and application of novel unsupervised learning methods, having in mind two target applications: the analysis of forensic case data and the classification of remote sensing images. First, a method based on a symbolic optimization of the inter-sample distance measure is proposed to improve the flexibility of spectral clustering algorithms, and applied to the problem of forensic case data. This distance is optimized using a loss function related to the preservation of neighborhood structure between the input space and the space of principal components, and solutions are found using genetic programming. Results are compared to a variety of state-of--the-art clustering algorithms. Subsequently, a new large-scale clustering method based on a joint optimization of feature extraction and classification is proposed and applied to various databases, including two hyperspectral remote sensing images. The algorithm makes uses of a functional model (e.g., a neural network) for clustering which is trained by stochastic gradient descent. Results indicate that such a technique can easily scale to huge databases, can avoid the so-called out-of-sample problem, and can compete with or even outperform existing clustering algorithms on both artificial data and real remote sensing images. This is verified on small databases as well as very large problems. Résumé : Ce travail de recherche porte sur le développement et l'application de méthodes d'apprentissage dites non supervisées. Les applications visées par ces méthodes sont l'analyse de données forensiques et la classification d'images hyperspectrales en télédétection. Dans un premier temps, une méthodologie de classification non supervisée fondée sur l'optimisation symbolique d'une mesure de distance inter-échantillons est proposée. Cette mesure est obtenue en optimisant une fonction de coût reliée à la préservation de la structure de voisinage d'un point entre l'espace des variables initiales et l'espace des composantes principales. Cette méthode est appliquée à l'analyse de données forensiques et comparée à un éventail de méthodes déjà existantes. En second lieu, une méthode fondée sur une optimisation conjointe des tâches de sélection de variables et de classification est implémentée dans un réseau de neurones et appliquée à diverses bases de données, dont deux images hyperspectrales. Le réseau de neurones est entraîné à l'aide d'un algorithme de gradient stochastique, ce qui rend cette technique applicable à des images de très haute résolution. Les résultats de l'application de cette dernière montrent que l'utilisation d'une telle technique permet de classifier de très grandes bases de données sans difficulté et donne des résultats avantageusement comparables aux méthodes existantes.

Mean first-passage time of continuous non-Markovian processes driven by colored noise

An equation for mean first-passage times of non-Markovian processes driven by colored noise is derived through an appropriate backward integro-differential equation. The equation is solved in a Bourret-like approximation. In a weak-noise bistable situation, non-Markovian effects are taken into account by an effective diffusion coefficient. In this situation, our results compare satisfactorily with other approaches and experimental data.

Dynamics of a paramagnetic colloidal particle driven on a magnetic-bubble lattice

We present a theoretical study of the recently observed dynamical regimes of paramagnetic colloidal particles externally driven above a regular lattice of magnetic bubbles [P. Tierno, T. H. Johansen, and T. M. Fischer, Phys. Rev. Lett. 99, 038303 (2007)]. An external precessing magnetic field alters the potential generated by the surface of the film in such a way to either drive the particle circularly around one bubble, ballistically through the array, or in triangular orbits on the interstitial regions between the bubbles. In the ballistic regime, we observe different trajectories performed by the particles phase locked with the external driving. Superdiffusive motion, which was experimentally found bridging the localized and delocalized dynamics, emerge only by introducing a certain degree of randomness into the bubbles size distribution.

Stochastic delay equations with non-negativity constraints driven by fractional Brownian motion

In this note we prove an existence and uniqueness result for the solution of multidimensional stochastic delay differential equations with normal reflection. The equations are driven by a fractional Brownian motion with Hurst parameter H > 1/2. The stochastic integral with respect to the fractional Brownian motion is a pathwise Riemann¿Stieltjes integral.

Stochastic delay differential equations driven by fractional Brownian motion with Hurst parameter H 1/2

We consider the Cauchy problem for a stochastic delay differential equation driven by a fractional Brownian motion with Hurst parameter H>¿. We prove an existence and uniqueness result for this problem, when the coefficients are sufficiently regular. Furthermore, if the diffusion coefficient is bounded away from zero and the coefficients are smooth functions with bounded derivatives of all orders, we prove that the law of the solution admits a smooth density with respect to Lebesgue measure on R.

Microcontroller-driven fluid-injection system for atomic force microscopy.

We present a programmable microcontroller-driven injection system for the exchange of imaging medium during atomic force microscopy. Using this low-noise system, high-resolution imaging can be performed during this process of injection without disturbance. This latter circumstance was exemplified by the online imaging of conformational changes in DNA molecules during the injection of anticancer drug into the fluid chamber.

Neutral and selection-driven decay of sexual traits in asexual stick insects.

Environmental shifts and lifestyle changes may result in formerly adaptive traits becoming non-functional or maladaptive. The subsequent decay of such traits highlights the importance of natural selection for adaptations, yet its causes have rarely been investigated. To study the fate of formerly adaptive traits after lifestyle changes, we evaluated sexual traits in five independently derived asexual lineages, including traits that are specific to males and therefore not exposed to selection. At least four of the asexual lineages retained the capacity to produce males that display normal courtship behaviours and are able to fertilize eggs of females from related sexual species. The maintenance of male traits may stem from pleiotropy, or from these traits only regressing via drift, which may require millions of years to generate phenotypic effects. By contrast, we found parallel decay of sexual traits in females. Asexual females produced altered airborne and contact signals, had modified sperm storage organs, and lost the ability to fertilize their eggs, impeding reversals to sexual reproduction. Female sexual traits were decayed even in recently derived asexuals, suggesting that trait changes following the evolution of asexuality, when they occur, proceed rapidly and are driven by selective processes rather than drift.

Local model predictive control experiences with differential driven wheeled mobile robots

This work extends a previously developed research concerning about the use of local model predictive control in differential driven mobile robots. Hence, experimental results are presented as a way to improve the methodology by considering aspects as trajectory accuracy and time performance. In this sense, the cost function and the prediction horizon are important aspects to be considered. The aim of the present work is to test the control method by measuring trajectory tracking accuracy and time performance. Moreover, strategies for the integration with perception system and path planning are briefly introduced. In this sense, monocular image data can be used to plan safety trajectories by using goal attraction potential fields

Phosphorus and carbon burial during cretaceous oceanic anoxic events, links with climatic changes

According to Jenkyns (2010), oceanic anoxic events (OAE) record profound changes in the climatic and paleoceanographic state of the planet and represent major disturbances in the global carbon cycle. One of the most studied OAEs on a worldwide scale is the Cenomanian-Turonian OAE 2, which is characterized by a pronounced positive excursion in carbon-isotope records and the important accumulation of organic-rich sediments. The section at Gongzha (Tibet) and the sections at Barranca and Axaxacualco (Mexico) are located in remote parts of the Tethys, and show δ13C records, which are well correlated with those of classical Tethyan sections. Both sections, however, do not exhibit the presence of organic-rich sediments. Phosphorus Mass Accumulation Rates (PMAR) in Tibet show a pattern similar to that observed in the Tethys by Mort et al. (2007), which suggests enhanced Ρ regeneration during the OAE 2 time interval, though there is no evidence for anoxic conditions in Tibet. Ρ appears here to have been mainly driven by detrital influx and sea-level fluctuations. The sections at Barranca and Axaxacualco show that the Mexican carbonate platform persisted during this anoxic event, which allowed the evolution of platform fauna otherwise not present in Tethyan sections. The persistence of this carbonate platform close to the Caribbean Igneous Plateau, which is thought to have released bio-limiting metals, is explained by local uplift which delayed the drowning of the platform and a specific oceanic circulation that permitted the preservation of oligotrophic conditions in the area. The Coniacian-Santonian OAE (OAE3) appears to have been more dependent on local conditions than OAE2. The presence of black shales associated with OAE3 appear to have been restricted to shallow-water settings and epicontinental seas in areas located around the Atlantic Ocean. The sections at Olazagutia (Spain), and Ten Mile - Arbor Park (USA), two potential Global Boundary Stratotype Sections and Points (GSSP) sites, are devoid of organic-rich sediments and lack a δ13C positive excursion around the C-S boundary. The Gabal Ekma section (Sinai, Egypt) exhibits accumulations of organic-rich sediments, in addition to phosphorite bone beds layers, which may have been linked to an epicontinental upwelling zone and/or storm inputs. Our data suggest that OAE 3 is rarely expressed by truly anoxic conditions and seems to have been linked to local conditions rather than global paleoenvironmental change. The evidence for detrital-P being the likely cause of Ρ fluctuations during the OAEs studied here does not negate the idea that anoxia was the principal driver of these fluctuations in the western Tethys. However, an explanation is required as to why the Ρ accumulation signatures are mirrored in both oxic and anoxic sedimentary successions. 'Eustatic/climatic' and 'productivity/anoxic' models may have both operated simultaneously in different parts of the world depending on local conditions, both producing similar trends in Ρ accumulation. - Selon Jenkyns (2010), les événements anoxiques océaniques enregistrent de profonds changements dans le climat et la paléoceanographie de la planète et représente des perturbations majeures du cycle du carbone. L'un des plus étudiés à l'échelle mondiale est l'ΟΑΕ2 du Cénomanien-Turonien, qui est caractérisé par une très forte excursion positive des isotopes du carbone et une importante accumulation de sédiments riche en matière organique. La section de Gongzha (Tibet) et les sections de Barranca et Axaxcualco (Mexique) sont situées aux confins de la Téthys, et enregistrent une courbe isotopique en δ13C parfaitement corrélable avec les sections téthysiennes, mais ne montre pas d'accumulation de black shales. Le taux de phosphore en accumulation de masses (PMAR) au Tibet montre un pattern similaire observé également par Mort et al. (2007) dans la Téthys, suggérant un model de régénération du Ρ durant l'anoxie, cependant aucune conditions anoxiques régnent dans la région du Tibet. Ρ apparaît donc principalement guidé par le détritisme et les fluctuations du niveau marin. Les sections de Barranca et d'Axaxacualco montrent que la plateforme carbonatée mexicaine persiste durant cet événement anoxique, et permet le développement d'une faune de plateforme qui n'est pas présente dans les sections téthysiennes. La persistance de cette plateforme carbonatée si proche du plateau Caribéen, qui est connu pour le relâchement de métaux bio-limitant, peut être expliqué par un soulèvement tectonique local qui inhibe l'ennoiement de la plateforme et une circulation océanique spécifique qui permet la préservation de conditions oligotrophiques dans cette région. L'événement anoxique océanique du Coniacien-Santonien apparaît plus dépendant des conditions locales que pour l'ΟΑΕ2. Les black shales associés à POAE3 sont restreints aux zones situées autour de l'océan Atlantique et plus particulièrement aux eaux peu profondes et épicontinentales. Les sections d'Olazagutia (Espagne), Ten Mile Creek et Arbor Park (USA), qui sont deux potentielles sections GSSP (Sections de stratotype de limite globaux et de points), ne montre pas d'accumulation de black shales et pas de forte excursion positive en δ13C autour de la limite C-S. La section de Gabal Ekma (Sinai, Egypte) montre des accumulations de black shales, en plus des couches de phosphorites et d'accumulation d'os (« bone beds »), vraisemblablement lié à des zones active d'upwelling épicontinentale et/ou d'apport de tempêtes. Nos données suggèrent que l'OAE 3 est rarement exprimé par de vraies conditions anoxiques et semble être plus lié à des conditions plus locales que des changements paléo-environnementaux globaux, comme observés pour le Cénomanien- Turonien. Les arguments pour un modèle lié au phosphore détritique qui serait la cause des fluctuations du phosphore total durant les OAEs, n'écartent pas l'idée que l'anoxie est la principale cause de ces fluctuations dans les sections riches en matière organique de l'Ouest téthysien. Cependant une explication est nécessaire pour comprendre pourquoi la signature de l'accumulation du phosphore est semblable dans les successions sédimentaires déposées dans des conditions oxygénées et anoxiques. Les modèles « Eustatisme/Climat » et « Productivité/anoxie » ont simultanément opéré dans les différentes parties du monde dépendant de conditions locales, et ont produit des tendances similaires en accumulation de phosphore.

Data-driven topo-climatic mapping with machine learning methods

Automatic environmental monitoring networks enforced by wireless communication technologies provide large and ever increasing volumes of data nowadays. The use of this information in natural hazard research is an important issue. Particularly useful for risk assessment and decision making are the spatial maps of hazard-related parameters produced from point observations and available auxiliary information. The purpose of this article is to present and explore the appropriate tools to process large amounts of available data and produce predictions at fine spatial scales. These are the algorithms of machine learning, which are aimed at non-parametric robust modelling of non-linear dependencies from empirical data. The computational efficiency of the data-driven methods allows producing the prediction maps in real time which makes them superior to physical models for the operational use in risk assessment and mitigation. Particularly, this situation encounters in spatial prediction of climatic variables (topo-climatic mapping). In complex topographies of the mountainous regions, the meteorological processes are highly influenced by the relief. The article shows how these relations, possibly regionalized and non-linear, can be modelled from data using the information from digital elevation models. The particular illustration of the developed methodology concerns the mapping of temperatures (including the situations of Föhn and temperature inversion) given the measurements taken from the Swiss meteorological monitoring network. The range of the methods used in the study includes data-driven feature selection, support vector algorithms and artificial neural networks.

A subcritical instability of wave-driven alongshore currents

The development of shear instabilities of a wave-driven alongshore current is investigated. In particular, we use weakly nonlinear theory to investigate the possibility that such instabilities, which have been observed at various sites on the U.S. coast and in the laboratory, can grow in linearly stable flows as a subcritical bifurcation by resonant triad interaction, as first suggested by Shrira eta/. [1997]. We examine a realistic longshore current profile and include the effects of eddy viscosity and bottom friction. We show that according to the weakly nonlinear theory, resonance is possible and that these linearly stable flows may exhibit explosive instabilities. We show that this phenomenon may occur also when there is only approximate resonance, which is more likely in nature. Furthermore, the size of the perturbation that is required to trigger the instability is shown in some circumstances to be consistent with the size of naturally occurring perturbations. Finally, we consider the differences between the present case examined and the more idealized case of Shrira et a/. [ 1997]. It is shown that there is a possibility of coupling between triads, due to the richer modal structure in more realistic flows, which may act to stabilize the flow and act against the development of subcritical bifurcations. Extensive numerical tests are called for.

A Smartphone-Based Prototype System for Incident/Work Zone Management Driven by Crowd-Sourced Data, 2015

This project develops a smartphone-based prototype system that supplements the 511 system to improve its dynamic traffic routing service to state highway users under non-recurrent congestion. This system will save considerable time to provide crucial traffic information and en-route assistance to travelers for them to avoid being trapped in traffic congestion due to accidents, work zones, hazards, or special events. It also creates a feedback loop between travelers and responsible agencies that enable the state to effectively collect, fuse, and analyze crowd-sourced data for next-gen transportation planning and management. This project can result in substantial economic savings (e.g. less traffic congestion, reduced fuel wastage and emissions) and safety benefits for the freight industry and society due to better dissemination of real-time traffic information by highway users. Such benefits will increase significantly in future with the expected increase in freight traffic on the network. The proposed system also has the flexibility to be integrated with various transportation management modules to assist state agencies to improve transportation services and daily operations.

GUS gene expression driven by a citrus promoter in transgenic tobacco and 'Valencia' sweet orange

The objective of this work was the transformation of tobacco and 'Valencia' sweet orange with the GUS gene driven by the citrus phenylalanine ammonia-lyase (PAL) gene promoter (CsPP). Transformation was accomplished by co-cultivation of tobacco and 'Valência' sweet orange explants with Agrobacterium tumefaciens containing the binary vector CsPP-GUS/2201. After plant transformation and regeneration, histochemical analyses using GUS staining revealed that CsPP promoter preferentially, but not exclusively, conferred gene expression in xylem tissues of tobacco. Weaker GUS staining was also detected throughout the petiole region in tobacco and citrus CsPP transgenic plants.

Gravity-driven slug motion in capillary tubes

The velocity of a liquid slug falling in a capillary tube is lower than predicted for Poiseuille flow due to presence of menisci, whose shapes are determined by the complex interplay of capillary, viscous, and gravitational forces. Due to the presence of menisci, a capillary pressure proportional to surface curvature acts on the slug and streamlines are bent close to the interface, resulting in enhanced viscous dissipation at the wedges. To determine the origin of drag-force increase relative to Poiseuille flow, we compute the force resultant acting on the slug by integrating Navier-Stokes equations over the liquid volume. Invoking relationships from differential geometry we demonstrate that the additional drag is due to viscous forces only and that no capillary drag of hydrodynamic origin exists (i.e., due to hydrodynamic deformation of the interface). Requiring that the force resultant is zero, we derive scaling laws for the steady velocity in the limit of small capillary numbers by estimating the leading order viscous dissipation in the different regions of the slug (i.e., the unperturbed Poiseuille-like bulk, the static menisci close to the tube axis and the dynamic regions close to the contact lines). Considering both partial and complete wetting, we find that the relationship between dimensionless velocity and weight is, in general, nonlinear. Whereas the relationship obtained for complete-wetting conditions is found in agreement with the experimental data of Bico and Quere [J. Bico and D. Quere, J. Colloid Interface Sci. 243, 262 (2001)], the scaling law under partial-wetting conditions is validated by numerical simulations performed with the Volume of Fluid method. The simulated steady velocities agree with the behavior predicted by the theoretical scaling laws in presence and in absence of static contact angle hysteresis. The numerical simulations suggest that wedge-flow dissipation alone cannot account for the entire additional drag and that the non-Poiseuille dissipation in the static menisci (not considered in previous studies) has to be considered for large contact angles.