Science: Image in Action

The following sections are included:

• Workshop Photographs

• ORGANIZING COMMITTEES

• Sponsors

• PREFACE

• Memory of Vito Di Gesù

• CONTENTS

We explore the concept of information in statistics: information about unknown quantities of interest, the parameters. We discuss intuitive ideas of what should be information in statistics. Our approach on information is divided in two scenarios: observed data and planning of an experiment. On the first scenario, we discuss the Sufficiency Principle, the Conditionality Principle, the Likelihood Principle and their relationship with trivial experiments. We also provide applications of some measures of information to an intuitive example. On the second scenario, the definition and new applications of Blackwell Sufficiency are presented. We discuss a new relationship between Blackwell Equivalence and the Likelihood Principle. Finally, the expected values of some measures of information are calculated.

We introduce a novel multi agent patrolling algorithm inspired by the behavior of gas filled balloons. Very low capability ant-like agents are considered with the task of patrolling an unknown area modeled as a graph. While executing the proposed algorithm, the agents dynamically partition the graph between them using simple local interactions, every agent assuming the responsibility for patrolling his subgraph. Balanced graph partition is an emergent behavior due to the local interactions between the agents in the swarm. Extensive simulations on various graphs (environments) showed that the average time to reach a balanced partition is linear with the graph size. The simulations yielded a convincing argument for conjecturing that if the graph being patrolled contains a balanced partition, the agents will find it. However, we could not prove this. Nevertheless, we have proved that if a balanced partition is reached, the maximum time lag between two successive visits to any vertex using the proposed strategy is at most twice the optimal so the patrol quality is at least half the optimal. In case of weighted graphs the patrol quality is at least of the optimal where l_max (l_min) is the longest (shortest) edge in the graph.

Noise plays a fundamental role in all living organisms from the earliest prokaryotes to advanced mammalian forms, such as ourselves. In the context of living organisms, the term 'noise' usually refers to the variance amongst measurements obtained from repeated identical experimental conditions, or from output signals from these systems. It is noteworthy that both these conditions are universally characterized by the presence of background fluctuations. In non-biological systems, such as electronics or in communications sciences, where the aim is to send error-free messages, noise was generally regarded as a problem. The discovery of Stochastic Resonances (SR) in non-linear dynamics brought a shift of perception where noise, rather than representing a problem, became fundamental to system function, especially so in biology. The question now is: to what extent is biological function dependent on random noise. Indeed, it seems feasible that noise also plays an important role in neuronal communication and oscillatory synchronization. Given this approach, it follows that determining Fisher information content could be relevant in neuronal communication. It also seems possible that the principle of least time, and that of the sum over histories, could be important basic principles in understanding the coherence dynamics responsible for action and perception. Ultimately, external noise cancellation combined with intrinsic noise signal embedding and, the use of the principle of least time may be considered an essential step in the organization of central nervous system (CNS) function.

Role of f-granulation in handling uncertainty in machine intelligence and its significance as a facet of natural computation are discussed. Its modeling through judicious integration of rough and fuzzy sets is explained. Several tasks like case generation, clustering, classification and segmentation are described in rough-fuzzy framework demonstrating the role of f-granulation and the resulting merits. A definition of generalized rough sets is given considering both the set and granules as crisp and/or fuzzy. Based on that new entropy measures are defined using exponential and logarithmic gain functions. The superiority of rough-fuzzy integration, in terms of performance and computation time, is illustrated for the tasks of case mining in large data sets, segmenting brain MR images, and classifying remotely sensed images as examples. Neighbourhood rough sets are capable of better dimensionality reduction. The effect of fuzzy-granules and generalizations in rough sets is demonstrated for image segmentation. The article includes some of the results published elsewhere.

The Baire metric induces an ultrametric on a dataset and is of linear computational complexity, contrasted with the standard quadratic time agglomerative hierarchical clustering algorithm. In this work we evaluate empirically this new approach to hierarchical clustering. We compare hierarchical clustering based on the Baire metric with (i) agglomerative hierarchical clustering, in terms of algorithm properties; (ii) generalized ultrametrics, in terms of definition. We apply the Baire distance to spectrometric and photometric redshifts from the Sloan Digital Sky Survey using, in this work, about half a million astronomical objects. We want to know how well the (more costly to determine) spectrometric redshifts can predict the (more easily obtained) photometric redshifts, i.e. we seek to regress the spectrometric on the photometric redshifts, and we use clusterwise regression for this.

Classification techniques are routinely utilized on satellite images. Pansharpening techniques can be used to provide super resolved multispectral images that can improve the performance of classification methods. So far, these pansharpening methods have been explored only as a preprocessing step. In this work we address the problem of adaptively modifying the pansharpening method in order to improve the precision and recall figures of merit of the classification of a given class without significantly deteriorating the performance of the classifier over the other classes. The validity of the proposed technique is demonstrated using a real Quickbird image.

Some blind approaches to a typical problem in astrophysical data analysis are briefly recalled. The problem is to disentangle different radiation maps from a multichannel map where signals coming from different astrophysical sources are mixed. Some separation methods assume a perfect knowledge of the mixing operator. However, if the model does not explain the data, this approach can lead to very poor results. Conversely, a blind approach, that is, one that does not enforce a rigid data model, is able to match the data while preserving all the desirable properties of the result.

The Orion Nebula and its associated young stellar cluster are located at the front-side of the optically thick OMC-1 molecular cloud. In order to disentangle the cluster members from background contamination, it is important to know the extinction provided by the OMC-1, which is poorly known, the available measurements yielding contradictory results. We analyze the most recent near-infrared catalog of stars to study the distribution of reddening across a 0.3 deg² area covering the Orion Nebula Cluster. On the basis of the observed photometry, we establish a statistical criterion for disentangling contaminants from bona-fide cluster members. For contaminant stars, interstellar reddenings are estimated by comparison with a synthetic galactic model. A statistical analysis is then performed to consistently account for local extinction, reddening and star-counts analysis. By means of this statistical approach, we derive the extinction map of the OMC-1 with angular resolution < 3′, deriving that the extinction provided by the OMC-1 is highly variable on spatial scales of a few arcminutes, while showing a general increase from the outskirts (A_V ~ 6) to the direction of the Trapezium asterism (A_V ≳ 30).

The present work focuses on the relation of physics, phenomenology and brain organization. Classical dualities behind these concepts are mapped into the difference between the physical notion of states and the accompanying notion of transitions. It is argued that physics does not strictly deal with the concept of transitions and a more complete description of nature and reality is feasible if aspects of transition phenomenology are added to the physical concept of states. In living organisms transitions between a certain set of states has a particularly important meaning. These transitions are associated with the phenomenon of experience. In this view the issue of contextuality is seen as a local change in the similarity dimension within a partition of some fractal phenomenal dynamics. I argue that involving this view into physics can help to resolve the problem behind physical realism and the 'tuning problem' between life and cosmology.

In this paper we shall address some meeting points between geometry and biology, in order to show that geometrical things and transformations take part intrinsically in the living systems. We focus on some features of macromolecular structures like DNA-proteins complexes. All things we speak about take place in the 3-dimensional space of a living cell and particularly in its nucleus, which of course interacts in many ways and at different levels with the whole cell, its cytoplasm and the organelles. Ideally, we think we should rather consider, instead of a 3-dimensional space, a configuration space characterized by all its phase spaces, since a living being is a very complex dynamical system, but this would be a too difficult, impossible task. This is of course a very partial view, an oversimplification, of what really happen in our organisms. Nevertheless, We believe that in biology we are today facing the following problem: how small or local changes in a living system do affect the global behaviour and response of the whole organisms? We search for an answer by arguing that mostly overall features of living systems are emergent properties of organization and regulation defined at the macroscopic level of their morphology and physiological behaviours, and also by showing that in complex living systems self-organization ensures robustness without loss of plasticity, in the sense that perturbations in the interactions properties of its single parts generally do not have damaging consequences on the living form as a whole.

This essay discusses a proposal that draws together the three great revolutionary theories of 20^th Century physics: quantum theory, relativity theory and chaos theory. Motivated by the Bohmian notion of implicate order, and what in chaos theory would be described as a strange attractor, the proposal attributes special ontological significance to certain non-computable, dynamically invariant state-space geometries for the universe as a whole. Studying the phenomenon of quantum interference, it is proposed to understand quantum wave-particle duality, and indeed classical electromagnetism, in terms of particles in space time and waves on this state space geometry. Studying the EPR experiment, the acausal constraints that this invariant geometry provides on spatially distant degrees of freedom, provides a way for the underlying dynamics to be consistent with the Bell theorem, yet be relativistically covariant ("nonlocality without nonlocality"). It is suggested that the physical basis for such non-computable geometries lies in properties of gravity with the information irreversibility implied by black hole no-hair theorems being crucial. In conclusion it is proposed that quantum theory may be emergent from an extended theory of gravity which is geometric not only in space time, but also in state space. Such a notion would undermine most current attempts to "quantise gravity".

This paper reviews some recent investigations of localized structures in one- and two-dimensional neural-like networks with complex threshold excitability.

In this paper, we propose a Generalized Explication of Causality (GEC) with an approach based both on an integral and on the local-causes point of view.

The language of Newton based on dynamical point of view of fluents and fluxions is the basic formalism (borrowed by analogy from Newton) for visualisation and quantitative description of the local-causes account.

We use the formalism of continuous media both from the Lagrangian and the Eulerian (local) point of view. The role of Observer in the definition of casual connection will be emphasized. The definition and the role of Extensive quantities in the GEC will be discussed.

All through the history of the universe there is an apparent tendency for increasing complexity, with the organization of matter in evermore elaborate and interactive systems. The living world in general, and the human brain in particular, provides the highest complexity known. It seems obvious that all of this complexity must be the result of physical, chemical and biological evolution, but it was only with Darwin that we began to get a scientific understanding of biological evolution. Darwinian principles are guiding in our understanding of such complex systems as the nervous system, but also for the evolution of human society and technology. Living organisms have to survive in a complex and changing environment. This implies response and adaption to environmental events and changes at several time scales. The interaction with the environment depends on the present state of the organism, as well as on previous experiences stored in its molecular and cellular structures. At a longer time scale, organisms can adapt to slow environmental changes, by storing information in the genetic material carried over from generation to generation. This phylogenetic learning is complemented by ontogenetic learning, which is adaptation at a shorter time scale, occuring in non-genetic structures. The evolution of a nervous system is a major transition in biological evolution and allows for an increasing capacity for information storage and processing, increasing chances of survival. Such neural knowledge processing, cognition, shows the same principal features as nonneural adaptive processes. Similarly, consciousness might appear, to different degrees, at different stages in evolution. Both cognition and consciousness depends critically on the organization and complexity of the organism. In this presentation, I will briefly discuss general principles for evolution of complexity, focussing on the evolution of the nervous system, which provides organisms with ever increasing capacity for complex behaviour, cognition and consciousness. I will also discuss some computational approaches, as tools for understanding relations between structure, dynamics and function of the nervous system.

Self-assembly is ubiquitous in physics, chemistry and biology, and has many applications in materials science and engineering. Here we present a general approach for finding the simplest set of building blocks that will assemble into a given physical structure. Our procedure can be adapted to any given geometry, and thus to any given type of physical system. The amount of information required to describe this simplest set of building blocks provides a quantitative measure of the structure's physical complexity, which is capable of detecting any symmetry or modularity in the underlying structure.We also introduce the notions of joint, mutual and conditional complexity for self-assembling structures. We illustrate our approach using self-assembling polyominoes, and demonstrate the breadth of its potential applications by using it to quantify the physical complexity of protein complexes.

The Category theory of Topological Thermodynamics and Continuous Topological Evolution can be used as an abstract, but universal, foundation for understanding, among other things, scale invariance, chaos, emergence, self-similarity, dynamical systems, irreversible processes, and other non-linear, and even statistical, phenomena. The abstract Category theory, as an exterior differential system, has solutions which can be put into correspondence with numerous physical experiments in different disciplines.

Spatio-temporal dynamics of the chain system modeling collective behavior of electrically coupled nonlinear cells is investigated. The dynamics of a local cell is described by two-component Morris–Lecar model. It is shown that such a system supports formation of two stable types of anti-phase spiking patterns (ASP). These are travelling and oscillatory type. High multistability of the travelling ASP is indicated.

This paper discusses relationships between reality, models and representations in the study of complex systems. We present three different case studies: the modeling of colliding galaxies, the modeling and simulation of human intelligence, the creation of digital avatars. We clarify the relationships between reality, models and representations using a semiotic approach. We show that visual representation is as important as the modeling and that there is no possible equivalence between the real phenomena and its model and representation.

In the past ten years, the concept of Virtual Observatory (VObs) has increasingly gained importance in the domain of astrophysics, as a way of seamlessly accessing data in different wavelength domains stored in digital archives. There are many reasons why the VObs is useful for the development of science: to monitor time variability of phenomena, to compare phenomena in different bands, to increase return for investment (by fostering data re-use for scientific, educational and outreach purposes), to perform statistical analysis and mining on large quantities of data. The International Virtual Observatory Alliance (IVOA) has paved the way for the VObs to become a really useful tool for the scientific community, by promoting standards, by defining data interoperability methods, by fostering the needed coordination among data providers. But the VObs is more than just archives and standards: it is also infrastructure, basic software tools, advanced applications, evolution of methods and techniques, cross-fertilization with other communities. Discovering information in wide-field images and mining large archives are key items towards the use of the VObs as a tool for developing science. Data mining, or knowledge discovery in databases, while being the main methodology to extract the scientific information contained in Massive Data Sets (MDS), needs to tackle crucial problems since it has to orchestrate complex challenges posed by transparent access to different computing environments, scalability of algorithms, reusability of resources. To achieve a leap forward for the progress of astrophysics in the data avalanche era, the community needs to implement an infrastructure capable of performing data access, processing and mining in a distributed but integrated context.

Nowadays, many scientific areas share the same broad requirements of being able to deal with massive and distributed datasets while, when possible, being integrated with services and applications. In order to solve the growing gap between the incremental generation of data and our understanding of it, it is required to know how to access, retrieve, analyze, mine and integrate data from disparate sources. One of the fundamental aspects of any new generation of data mining software tool or package which really wants to become a service for the community is the possibility to use it within complex workflows which each user can fine tune in order to match the specific demands of his scientific goal. These workflows need often to access different resources (data, providers, computing facilities and packages) and require a strict interoperability on (at least) the client side. The project DAME (DAta Mining & Exploration) arises from these requirements by providing a distributed WEB-based data mining infrastructure specialized on Massive Data Sets exploration with Soft Computing and machine learning methods. It results as a multi-disciplinary platform-independent tool perfectly compliant with modern KDD (Knowledge Discovery in Databases) requirements and Information & Communication Technology trends.

Galaxies are argued to manifest complexity, thereby contradicting models of smooth parametric galactic phase space densities. An estimation of chaos in models of our galaxy is forwarded to suggest strength and possible causes of non-linearities in phase space. A Bayesian nonparametric methodology that is designed to acknowledge uncertainties in measured data, when applied to an observed external galaxy, indicates a non-linear galactic phase space density that could result from a bistable system potential. The effect of such a phase space structure on the inverse modelling of phase space data is discussed.

Fluorescence tomography is treated as a shape reconstruction problem for a coupled system of two linear transport equations in 2D. The shape evolution is designed in order to minimize the least squares data misfit cost functional either in the excitation frequency or in the emission frequency. Furthermore, a level set technique is employed for numerically modelling the evolving shapes. Numerical results are presented which demonstrate the performance of this novel technique in the situation of noisy simulated data in 2D.

Symmetry plays a fundamental role in aiding the visual system, to organize its environmental stimuli and to detect visual patterns of natural and artificial objects. Various kinds of symmetry exist, and we will discuss how internal symmetry due to textures influences the choice of direction in visual tasks. Two experiments are presented: the first, with human subjects, deals with the effect of textures on preferences for a pointing direction. The second emulates the performances obtained in the first through the use of an algorithm based on a physic metaphor. Results from both experiments are shown and comment.

In recent years many procedures have been proposed to check the anisotropy of a dataset. We present a new simple procedure, based on a scale dependent approach, to detect anisotropy signatures in a given distribution with particular attention to small dataset. The method provides a good discrimination power for both large and small datasets, even in presence of strong contaminating isotropic background. We present some applications to simulated datasets of events to investigate statistical features of the method and present and inspect its behavior under both the null or the alternative hypothesis.

This paper presents a new method for the identification of extensive air showers initiated by different primaries. The method uses the multiscale concept and is based on the analysis of multifractal behaviour and lacunarity of secondary particle distributions together with a properly designed and trained artificial neural network. The separation technique is particularly suited for being applied when the topology of the particle distribution in the shower front is as largely detailed as possible. In the present work the method is discussed and applied in the experimental framework of ARGO-YBJ, to obtain hadron to gamma primary separation. We show that the presented approach gives very good results, leading, in the 1 – 10 Tev energy range, to a clear improvement of the discrimination power with respect to the existing figures for extended shower detectors.

The Bayesian semi-parametric curve-fitting procedure, based on a new, flexible, fast, and efficient mixture analysis idea assuming unknown number of components is applied to analyze SDSS data to study the relationship between apparent magnitude and redshift for quasars and the possibility of clusterings. The cosmological data analysis provides strong evidence against linear relationship, and clearly indicates the possibility of clustering of quasars specially at high redshift. This sheds new light not only on the issue of evolution, existence of acceleration or deceleration and environment around quasars (say, radio loud and radio quiet) at high redshift but also help us to estimate the cosmological parameters related to acceleration or decceleration.

The following sections are included:

• AUTHOR INDEX

• PARTICIPANTS