Skip to main content

Über dieses Buch

The following are the proceedings of the Fourth International Workshop on Human and Machine Perception held in Palermo, Italy, on June 20 -23, 2000, under the auspices of three Institutions: the Cybernetic and Biophysics Group (GNCB) of the Italian National Research Council (CNR) and the two Inter-Department Centers of Cognitive Sciences of Palermo and Pavia University respectively. A broad spectrum of topics are covered in this series, ranging from computer perception to psychology and physiology of perception. The theme of this workshop on Human and Machine Perception was focused on Thinking, Deciding, and Acting. As in the past editions the final goal has been the analysis and the comparison of biological and artificial solutions. The focus of the lectures has been on presenting the state-of-the-art and outlining open questions. In particular, they sought to stress links, suggesting possible synergies between the different cultural areas. The panel discussion has been conceived as a forum for an open debate, briefly introduced by each panelist, and mainly aimed at deeper investigation of the different approaches to perception and strictly related topics. The panelists were asked to prepare a few statements on hot-points as a guide for discussion. These statements were delivered to the participants together with the final program, for a more qualified discussion.



General Talks

From Computing With Numbers to Computing With Words - From Manipulation of Measurements to Manipulation of Perceptions

Computing, in its usual sense, is centered on manipulation of numbers and symbols. Incontrast, computing with words, or CW for short, is a methodology in which the objects of computation are words and propositions drawn from a natural language, e.g., small, large,far, heavy, not very likely, the price of gas is low and declining, Berkeley is near San Francisco, it is very unlikely that there will be a significant increase in the price of oil in the near future, etc. Computing with words is inspired by the remarkable human capability to perform a wide variety of physical and mental tasks without any measurements and any computations. Familiar examples of such tasks are parking a car, driving in heavy traffic,playing golf, riding a bicycle, understanding speech and summarizing a story. Underlying this remarkable capability is the brain’s crucial ability to manipulate perceptions- perceptions of distance, size, weight, color, speed, time, direction, force, number, truth,likelihood and other characteristics of physical and mental objects. Manipulation of perceptions plays a key role in human recognition, decision and execution processes. As a methodology, computing with words provides a foundation for a computational theory of perceptions - a theory which may have an important bearing on how humans make – and machines might make - perception-based rational decisions in an environment of imprecision, uncertainty and partial truth.
A basic difference between perceptions and measurements is that, in general,measurements are crisp whereas perceptions are fuzzy. One of the fundamental aims of science has been and continues to be that of progressing from perceptions to measurements.Pursuit of this aim has led to brilliant successes. We have sent men to the moon; we can build computers that are capable of performing billions of computations per second; we have constructed telescopes that can explore the far reaches of the universe; and we can date the age of rocks that are millions of years old. But alongside the brilliant successes stand conspicuous underachievements and outright failures. We cannot build robots which can move with the agility of animals or humans; we cannot automate driving in heavy traffic;we cannot translate from one language to another at the level of a human interpreter; we cannot create programs which can summarize non-trivial stories; our ability to model the behavior of economic systems leaves much to be desired; and we cannot build machines that can compete with children in the performance of a wide variety of physical and cognitive tasks.
It may be argued that underlying the underachievements and failures is the unavailability of a methodology for reasoning and computing with perceptions rather than measurements. An outline of such a methodology - referred to as a computational theory of perceptions - is presented in this paper. The computational theory of perceptions, or CTP for short, is based on the methodology of computing with words (CW). In CTP, words play the role of labels of perceptions and, more generally, perceptions are expressed as propositions in a natural language. CW-based techniques are employed to translate propositions expressed in a natural language into what is called the Generalized Constraint Language(GCL). In this language, the meaning of a proposition is expressed as a generalized constraint, X isr R, where X is the constrained variable, R is the constraining relation and isr is a variable copula in which r is a variable whose value defines the way in which R constrains X. Among the basic types of constraints are:possibilistic,veristic,probabilistic,random set, Pawlak set, fuzzy graph and usuality. The wide variety of constraints in GCL makes GCL a much more expressive language than the language of predicate logic.
In CW, the initial and terminal data sets, IDS and TDS, are assumed to consist of propositions expressed in a natural language. These propositions are translated, respectively,into antecedent and consequent constraints. Consequent constraints are derived from antecedent constraints through the use of rules of constraint propagation. The principal constraint propagation rule is the generalized extension principle. The derived constraints are retranslated into a natural language, yielding the terminal data set (TDS). The rules of constraint propagation in CW coincide with the rules of inference in fuzzy logic. A basic problem in CW is that of explicitation of X, R and r in a generalized constraint, X isr R,which represents the meaning of a proposition, p, in a natural language.
There are two major imperatives for computing with words. First, computing with words is a necessity when the available information is too imprecise to justify the use of numbers; and second, when there is a tolerance for imprecision which can be exploited to achieve tractability, robustness, low solution cost and better rapport with reality. Exploitation of the tolerance for imprecision is an issue of central importance in CW and CTP. At this juncture, the computational theory of perceptions - which is based on CW – is in its initial stages of development. In time, it may come to play an important role in the conception,design and utilization of information/ intelligent systems. The role model for CW and CTP is the human mind.
Lotfi A. Zadeh

Processing of Biological Signals: A Bridge Between Structure and Function

A biological structure is the source of a large number of signals of different nature, among which we may recall the electroencephalogram, the evoked and event related brain potentials, the electrocardiogram, the electromyogram, blood pressures (recorded in different body sites), fluid flows, biochemical concentrations, and many others. Most of them can be obtained non invasively and all of them carry information about the considered body organ or, from a more general point of view, about the whole signal generating process, that sometimes involves complex structures and systems. Due to this complexity the informative contents have sometimes to be enhanced through advanced signal processing techniques able to eliminate the superimposed noise, to extract new parameters, to quantitatively relate different signals. In particular, multiparametric approaches may lead to the definition of mathematical models of the underlying structure, for facilitating in the interpretation both in physiological and pathological conditions.
Sergio Cerutti, Anna M. Bianchi, Maria G. Signorini

Architectures for Artificial Mentation and Behavior

This work supports the hypothesis that behavior is the result of rich but mediated representations of the real world and the organism within it. ‘Rich’ means containing all the detail necessary for appropriate action taking and ‘mediated’ means that representations of ’self are at work which include decisions on whether and how to act. The central issue is an understanding of the emergent depictive properties of brain-inspired complex architectures of which the components are neural network modules each with their own emergent properties. This work contrasts markedly with Artificial Intelligence techniques, a point which is made early in the chapter.
Igor Aleksander

Developing Rational Agents

An agent is a system that interacts with an environment continually and without human assistance in order to carry out a predefined task. We are interested in developing artificial agents that act rationally, in the sense that they are able to maximize a suitable utility function. In this chapter, we describe the main problems underlying the realization of rational agents and present commonly adopted mathematical models. In particular, we consider the case in which the environment can be modeled as a finite state stochastic process and address the problem of developing agents that can learn to act rationally through their own experience.
Marco Colombetti, Pier Luca Lanzi

Overview on Decision Making

The systematic empirical study of judgement and decision making did not begin to emerge as a discipline in its own right until the 1960s, when there was an upsurge of interest in the broader and more general field of cognitive psychology that includes memory, thinking, problem solving, mental imagery and language.
Giuseppe Mosconi, Laura Macchi

Panel Summary: Think About This: is Thinking Necessary to do Actions?

When Virginio Cantoni, Vito Di Gesù, and one of the authors (JDB), in a very creative meeting in Palermo, decided about the title of this panel discussion, the original formulation simply was: Is thinking necessary?
Jean Beaudrillard has argued that in the future thinking will be a useless luxury. Josef Lechner has stated that there is a complementarity between thinking and happiness. Niko Tinbergen has conjectured that intelligence in man has developed for the ability to predict the actions of other people to a higher degree (so here we already have a connection between thinking and action), and that thinking about general relativity or artificial intelligence is an abuse of the brain. All these statements are not very encouraging.
Jörg D. Becker, Mariagrazia Semenza, Alessandra Setti


Learning from Mistakes

Learning from mistakes is probably the most familiar form of learning for every human being. Every time we make a mistake, i.e., we suffer from an inconvenient due to our behavior, we try to modify our knowledge about the world in order to avoid suffering again in the future for the same reason. However, human learning is something of very complex and still poorly understood, so that it is hard to say which are the mechanisms triggered by the detection of a mistake. Even more difficult is to relate this form of learning to other human typical forms of learning, such as learning from textbooks.
Attilio Giordana, Alessandro Serra

Panel Summary: the Parallel Between Acting and Thinking: Using it for Better Process Understanding

The aim of the panel session was to discuss the couple ACTING/THINKING in both living and manufactured systems. Trying to avoid semantic and semiotic discussions on what thinking really means, we proposed two hypotheses to start:
  • thinking deals with building and transforming representations;
  • there is a fruitful parallel to draw between thinking and acting.
We based and evaluated them on progressively more abstract examples recalled here. There were many interventions, comments and objections, during the introductory statement and in the subsequent discussion: they are reported in the last section.
Bertrand Zavidovique

Time-Effective Detection of Objects of Interest in Images by Means of A Visual Attention Mechanism

Time-effective detection and recognition of objects of interest in images is still a matter of intensive research in computer vision community because the artificial vision systems fail to outperform the detection results by a human being. The detection problem is complicated when objects of interest have low contrast and various sizes or orientations and can be located on noisy and inhomogeneous background with occlusion occurrences. In many practical applications, the real-time implementation of object detection algorithms in such conditions is a matter of great concern. The results of numerous neurological and psychophysical investigation of human visual system (HVS) indicate that the human vision can successfully cope with these complex situations because of using the visual attention mechanism associated with a model-based image analysis.1,2 The goal of the investigation presented here was not the simulation of human visual perception but the incorporation of its advantageous features into computer vision algorithms. Besides many remarkable properties of HVS like the mentioned model-based visual attention, the HVS has also some disadvantages while detecting and identifying objects. For instance, it is prone to illusions that should be not automatically copied onto an artificial vision system.3
Roman M. Palenichka, Peter Zinterhof

Experiments on Concurrent Artificial Environment

We show how the simulation of concurrent system is of interest for both behavioral studies and strategies of learning applied on prey-predator problems. In our case learning studies into unknown environment have been applied to mobile units by using genetic algorithms (GA). A set of trajectories, generated by GA, are able to build a description of the external scene driving a predators to a prey. Here, an example of prey-predator strategy,based on field of forces, is proposed. The evolution of the corespondent system can be formalized as an optimization problem and, for that purpose, GA can be use to give the right solution at this problem. This approach could be applied to the autonomous robot navigation in risky or inaccessible environments (monitoring of atomic power plants, exploration of sea bottom, and space missions).
Vito Di Gesù, Giosuè Lo Bosco, Domenico Tegolo

Rational Decision: A Matter of Common Sense

All of us have to make decisions every day of our lives. Many empirical studies have been made which required extensive surveys and considerable laboratory work, including investigations on simple decision-making processes. They mostly appear to show that human beings are incoherent.
Mariagrazia Semenza

A Hierarchical Entropy Based Representation for Medical Signals

In this chapter, we present a Hierarchical Entropy-based Representation (HER) for 1-dimensional signals. In fact, many 2-dimensional signals can be effectively encoded in a 1-D form that will allow the use of HER. This method represents the signal by a vector containing the energy values related to its local maxima and their locations along the time axis. Such a representation has been utilized as the feature extraction engine in an image retrieval system for medical image databases, with a K-d-tree based spatial access method.The data used for the experiments were contours and textures from various medical sources.The experiments show that HER performs very well in this respect.
Cecilia Di Ruberto, Riccardo Distasi, Sergio Vitulano

Adaptation and Learning for Pattern Recognition: A Comparison Between Neural and Evolutionary Computation

The work presented in this chapter is aimed at developing self-governing artificial systems that are able to operate in complex, uncertain and dynamic application domains, by mimicking the learning and adaptation capabilities exhibited by biological systems. For this purpose, we are exploring the possibility offered by the artificial neural networks and evolutionary computation paradigms for automatically extracting the set of prototypes describing the variability present in a data set. In particular, this chapter reports the results of an experiment designed for comparing the performance exhibited by a Learning Vector Quantization network and an Evolutionary Learning System using a Breeder Genetic Algorithm. For the sake of generality, the comparison has been performed on a complex classification problem obtained by generating a synthetic data set according to the distribution of distributions model.
Claudio De Stefano, Antonio Della Cioppa, Angelo Marcelli


Decision Making in Evolving Artificial Systems

The theme of this workshop is artificial perception. In this chapter we will argue that the ecological function of perception is to serve decision-making. If this is so the mechanisms chosen to implement perception, in natural or artificial systems, will be constrained by the requirements of decision-making and theories of decision-making will inevitably influence theories of perception. In what follows we will look at decision-making from what we hope is a new perspective, applying concepts and techniques developed by what we will call “new artificial intelligence”. We will begin, in the second part of the chapter, with a review of traditional, “normative” theories of decision-making and of the mounting body of experimental evidence, showing the inadequacy of these theories. In the third part we will outline an alternative ecological/biological view of decision-making, inspired by computer models from the fields of “Artificial Neural Networks”, “Artificial Evolution” and “Evolutionary Robotics. Finally, we will briefly examine how far this new view may be considered to be biologically realistic and examine some of the possible consequences for theories of perception and of individual perceptive systems.
Richard Walker, Maurizio Cardaci

Case Studies in Cognitive Robotics

Cognitive Robotics is concerned with the theory and implementation of embodied intelligent agents. Cognitive Robotics poses a unique challenge to researchers: it deals in fact with manufacts that are inherently autonomous, situated in a not structured environment, and whose behavior is characterized by a cycle “Perception-Reasoning-Action” or “Sense-Plan-Act” (sometimes called SPA architectures). Of course this cycle has to be performed reliably, and in real time. Sensing, planning and acting are difficult problems to be solved, per se. Their integration within a single system poses additional challenges.
Luigia Carlucci Aiello, Daniele Nardi, Fiora Pirri

Panel Summary: Rational and Not Rational Behaviors in Decision-Making

The aim of this panel is to address the complexity of deciding as it is stated in the workshop agenda which says, as it is known: Decision Theory provides a methodology for dealing with such circumstances. Moreover, is decision a simple consequence of reasoning on facts ? This simple question put several drawbacks concerning the decision level (hierarchical view), its complexity and representation.
Roberto Bordogna, Vito Di Gesù, Bertrand Zavidovique, Virginio Cantoni


Biological Brain and Mind

It is particularly appropriate to speak in these days about brain, that is biological brain, and its relationship with mind. A lot has been learnt about this topic in the past 20 years or so. This is due primarily to the convergence of three powerful research lines: experimental psychology (nowadays often termed cognitive science), biology (especially molecular biology) and brain imaging.
Edoardo Boncinelli

Visual Attention Systems

The attention processes of vision systems are thought-out outlining some analogies between biological vision and the solutions given by computer scientists in artificial vision.The studies on human visual attention identify two phases: a pre-attentive one in which the visual system detects regions of interest within the full field of view, also through simple and fast alerting mechanisms, and an attentive one in which each detected region is analyzed in detail. The same computational paradigm is considered for artificial vision systems in order to successfully deal with the enormous quantity of raw data transduced by standard vision sensors. Such a paradigm exploits the variable resolution grids, according to the image detail required for the task at hand, so fully utilizing the capabilities of multi-resolution and pyramid computer vision systems.
Virginio Cantoni, Massimo Cellario

Panel Summary: Planning And Plasticity in Artificial and Natural Systems

Boncinelli underlined often, during our workshop, that simplest organisms are the more robust and they survived longer than more evolved organisms (what about the quality of their life?). This depends on the fact that complex systems either natural or artificial need for more complex controls. Moreover, planning strategies, one of the steps for finding optimal solutions, must be carefully designed and it is complex too.
Vito Di Gesù, Settimo Termini, Virginio Cantoni, Bertrand Zavidovique


Notion Formation in Machine Learning

In Artificial Intelligence, and, specifically in Machine Learning, to form a new notion usually means to build up a “concept” or a “category”. The simplest way to consider a category is extensional: a category is a set of “equivalent” objects,1 i.e., objects that share properties.2 Categories are organized into taxonomies, linked through the set inclusion relation.1 A concept is sometime considered as the intensional representation of a category, i.e., a description of the objects in the category,2 called also the instances of the concept. Actually, most frequently, the denotations “concept” and “category” are used as synonyms.
Lorenza Saitta, Roberto Esposito

Towards Artificial Action: Teaching by Showing

This paper describes an approach to using machine vision to provide sensor feedback to robotic manipulators and shows how tasks may be taught, represented and repeated. This work was undertaken in order to increase the flexibility within a plant in which robotic manipulators are used to weld ship parts. This task has several features which are characteristic of a broad range of tasks in flexible manufacturing; the robot is required to accurately position a tool relative to a target structure (the workpiece) which may be inaccurately located with respect to the robot. This requirement imposes the constraint that the robot must be able to sense the structures in its environment in some way and machine vision provides a flexible way to achieve this.
T. Drummond, R. Cipolla

Panel Summary: Frontiers of Human-Machine Interaction

The hot points presented to the panel were the following:
  • What are the broader definitions of Human/Machine interaction? For example:
    • non direct connection (using currently available computer GUI)
    • first phase direct connection (linking sensors to the sensory system)
    • second phase (direct connections into the Central Nervous System)
  • What are the technologies that should be developed in order to enable each of the previous phases?
  • What are the scientific research issues that are related to such phases?
  • What are the possible implications of a direct human-computer link on the society? (compared to the Internet revolution?)
  • There are several phases in Human-Machine interaction. The first, and most trivial one, involves using virtual reality and augmented reality devices, that allow even today manipulation of objects by interactively superpositioning artificial objects and data on real images- The next interesting phases, however, involves direct data transfer from artificial sensors to the nervous system. In that regard, we can show various implementations of artificial auditory and visual sensors. Hearing implants are being coupled into the auditory system, and a direct link to the auditory nerve does exist. Regarding visual data, the task is more complicated, and current technologies are being developed in order to implant a sensor into the retina and connect it to the optic nerve.
Gianni Rigamonti, Nathan Intrator, Hezy Yeshurun


Weitere Informationen