A Theory of Networks for Appxoimation and Learning

Learning an input-output mapping from a set of examples, of the type that many neural networks have been constructed to perform, can be regarded as synthesizing an approximation of a multi-dimensional function, that is solving the problem of hypersurface reconstruction. From this point of view, this form of learning is closely related to classical approximation techniques, such as generalized splines and regularization theory. This paper considers the problems of an exact representation and, in more detail, of the approximation of linear and nolinear mappings in terms of simpler functions of fewer variables. Kolmogorov's theorem concerning the representation of functions of several variables in terms of functions of one variable turns out to be almost irrelevant in the context of networks for learning. We develop a theoretical framework for approximation based on regularization techniques that leads to a class of three-layer networks that we call Generalized Radial Basis Functions (GRBF), since they are mathematically related to the well-known Radial Basis Functions, mainly used for strict interpolation tasks. GRBF networks are not only equivalent to generalized splines, but are also closely related to pattern recognition methods such as Parzen windows and potential functions and to several neural network algorithms, such as Kanerva's associative memory, backpropagation and Kohonen's topology preserving map. They also have an interesting interpretation in terms of prototypes that are synthesized and optimally combined during the learning stage. The paper introduces several extensions and applications of the technique and discusses intriguing analogies with neurobiological data.

Abordagem neuropsicológica das alterações na atenção e memória no sindroma de Burnout

O sindroma de Burnout, quadro psicofisio-patológico tem sido objecto de investigação intensiva, desde o artigo de Freudenberger (1974) intitulado "Staff Burnout", com dois objectivos: compreendê-lo melhor, através de meios de diagnóstico, e criar técnicas de intervenção terapêutica. Na realidade, desde essa altura, foram efectuados e publicados um número avultado de trabalhos de investigação, nos campos do diagnóstico e caracterização do Burnout, e da sua resolução terapêutica. O pensamento dominante, nessa altura e ainda hoje, é de tendência analítica e/ou psico-social. Este quadro, espoletado por uma sucessão de episódios emocionalmente negativos em contexto ocupacional em indivíduos com provável predisposição genética e sujeitos a situações de pressão laboral, dos mais diversos tipos (podendo ir do “simples” stress por acumulação de tarefas até às situações de mobbing), tem efeitos frequentemente dramáticos ao nível da dinâmica biopsico- social, nos seus mais diversos aspectos. Estes estendem-se, quase sempre, muito para lá das problemáticas laborais, prejudicando, de forma mais ou menos grave, as interacções sociais com particular impacto ao nível da dinâmica familiar. Por outro lado, o Burnout propicia o aparecimento de patologias diversas, já que toda a estrutura psiconeuro-endocrino-imunulógica estará posta em causa, potenciando situações de fragilidade sistémica. No entanto, há aspectos correlacionáveis com este quadro disfuncional que têm sido muito pouco abordados – alterações cognitivo-operativas ou neuropsicológicas. Aliás os trabalhos que sobre eles incidem são em número muito reduzido. Assim após termos registado queixas, acentuadas, ao nível da capacidade de concentração e da memória em pessoas com burnout observadas na clínica hospitalar e privada, decidimos investigar estas situações, usando uma metodologia clínica de tipo qualitativo, e constatámos que, na realidade, as queixas eram pertinentes. Posto isto, achámos que a situação deveria ser aprofundada e partimos para um trabalho mais sistematizado, este, com o objectivo de caracterizar melhor o tipo de disfunções atencionais e mnésicas. Para isso, após uma selecção prévia, a partir de um grupo de 192 enfermeiros que responderam à Escala de Maslach, avaliámos uma amostra de risco constituída por 40 enfermeiros e enfermeiras, de Instituições Psiquiátricas da Grande Lisboa, trabalhando em urgência e enfermaria, que comparámos com uma amostra de igual número de enfermeiros, desenvolvendo a sua actividade na consulta externa ou em ambientes mais protegidos de stress ocupacional continuo. Para o efeito, e após uma anamnese cuidada, aplicámos provas de atenção e memória, sensíveis a qualquer tipo de compromisso encefálico seja ele funcional ou patológico. Para a componente atenção/concentração e a componente vísuo-grafo-espacial usámos a prova de Toulouse-Piéron, assim como as séries de dígitos ou digit span, para a vertente audio-verbal. A dinâmica mnésica foi avaliada através da prova de memória associativa (Escala de Memória de Wechsler) para testar a variante áudio-verbal, e a reprodução de figuras (Escala de Memória de Wechlser). Os resultados, após uma dupla análise clínica e estatística, comprovaram globalmente as hipóteses, indicando uma correlação significativa entre o grau de Burnout e os défices neuropsicológicos detectados: alteração da atenção/concentração e dismnésia, de natureza limitativa face às exigências quotidianas dos indivíduos. Finalmente, com base na revisão da literatura e os resultados deste estudo, foi esquematizado um Modelo Neuropsicológico do sindroma de Burnout, que nos parece espelhar as relações entre este quadro clínico, as alterações cognitivooperativas encontradas e as principais estruturas encefálicas, que julgamos, implicadas em toda a dinâmica do processo disfuncional.

Self-organising neural networks for adaptive control

Self-organizing neural networks have been implemented in a wide range of application areas such as speech processing, image processing, optimization and robotics. Recent variations to the basic model proposed by the authors enable it to order state space using a subset of the input vector and to apply a local adaptation procedure that does not rely on a predefined test duration limit. Both these variations have been incorporated into a new feature map architecture that forms an integral part of an Hybrid Learning System (HLS) based on a genetic-based classifier system. Problems are represented within HLS as objects characterized by environmental features. Objects controlled by the system have preset targets set against a subset of their features. The system's objective is to achieve these targets by evolving a behavioural repertoire that efficiently explores and exploits the problem environment. Feature maps encode two types of knowledge within HLS — long-term memory traces of useful regularities within the environment and the classifier performance data calibrated against an object's feature states and targets. Self-organization of these networks constitutes non-genetic-based (experience-driven) learning within HLS. This paper presents a description of the HLS architecture and an analysis of the modified feature map implementing associative memory. Initial results are presented that demonstrate the behaviour of the system on a simple control task.

Neurofuzzy mixture of experts network parallel learning and model construction algorithms

A connection between a fuzzy neural network model with the mixture of experts network (MEN) modelling approach is established. Based on this linkage, two new neuro-fuzzy MEN construction algorithms are proposed to overcome the curse of dimensionality that is inherent in the majority of associative memory networks and/or other rule based systems. The first construction algorithm employs a function selection manager module in an MEN system. The second construction algorithm is based on a new parallel learning algorithm in which each model rule is trained independently, for which the parameter convergence property of the new learning method is established. As with the first approach, an expert selection criterion is utilised in this algorithm. These two construction methods are equivalent in their effectiveness in overcoming the curse of dimensionality by reducing the dimensionality of the regression vector, but the latter has the additional computational advantage of parallel processing. The proposed algorithms are analysed for effectiveness followed by numerical examples to illustrate their efficacy for some difficult data based modelling problems.

Neurofuzzy design and model construction of nonlinear dynamical processes from data

A common problem in many data based modelling algorithms such as associative memory networks is the problem of the curse of dimensionality. In this paper, a new two-stage neurofuzzy system design and construction algorithm (NeuDeC) for nonlinear dynamical processes is introduced to effectively tackle this problem. A new simple preprocessing method is initially derived and applied to reduce the rule base, followed by a fine model detection process based on the reduced rule set by using forward orthogonal least squares model structure detection. In both stages, new A-optimality experimental design-based criteria we used. In the preprocessing stage, a lower bound of the A-optimality design criterion is derived and applied as a subset selection metric, but in the later stage, the A-optimality design criterion is incorporated into a new composite cost function that minimises model prediction error as well as penalises the model parameter variance. The utilisation of NeuDeC leads to unbiased model parameters with low parameter variance and the additional benefit of a parsimonious model structure. Numerical examples are included to demonstrate the effectiveness of this new modelling approach for high dimensional inputs.

Optimal piecewise locally linear modeling

Associative memory networks such as Radial Basis Functions, Neurofuzzy and Fuzzy Logic used for modelling nonlinear processes suffer from the curse of dimensionality (COD), in that as the input dimension increases the parameterization, computation cost, training data requirements, etc. increase exponentially. Here a new algorithm is introduced for the construction of a Delaunay input space partitioned optimal piecewise locally linear models to overcome the COD as well as generate locally linear models directly amenable to linear control and estimation algorithms. The training of the model is configured as a new mixture of experts network with a new fast decision rule derived using convex set theory. A very fast simulated reannealing (VFSR) algorithm is utilized to search a global optimal solution of the Delaunay input space partition. A benchmark non-linear time series is used to demonstrate the new approach.

Data based constructive identification - overcoming the curse of dimensionality

The modelling of a nonlinear stochastic dynamical processes from data involves solving the problems of data gathering, preprocessing, model architecture selection, learning or adaptation, parametric evaluation and model validation. For a given model architecture such as associative memory networks, a common problem in non-linear modelling is the problem of "the curse of dimensionality". A series of complementary data based constructive identification schemes, mainly based on but not limited to an operating point dependent fuzzy models, are introduced in this paper with the aim to overcome the curse of dimensionality. These include (i) a mixture of experts algorithm based on a forward constrained regression algorithm; (ii) an inherent parsimonious delaunay input space partition based piecewise local lineal modelling concept; (iii) a neurofuzzy model constructive approach based on forward orthogonal least squares and optimal experimental design and finally (iv) the neurofuzzy model construction algorithm based on basis functions that are Bézier Bernstein polynomial functions and the additive decomposition. Illustrative examples demonstrate their applicability, showing that the final major hurdle in data based modelling has almost been removed.

Both younger and older adults have difficulty updating emotional memories

Objective. The main purpose of the study was to examine whether emotion impairs associative memory for previously seen items in older adults, as previously observed in younger adults. Method. Thirty-two younger adults and 32 older adults participated. The experiment consisted of 2 parts. In Part 1, participants learned picture–object associations for negative and neutral pictures. In Part 2, they learned picture–location associations for negative and neutral pictures; half of these pictures were seen in Part 1 whereas the other half were new. The dependent measure was how many locations of negative versus neutral items in the new versus old categories participants remembered in Part 2. Results. Both groups had more difficulty learning the locations of old negative pictures than of new negative pictures. However, this pattern was not observed for neutral items. Discussion. Despite the fact that older adults showed overall decline in associative memory, the impairing effect of emotion on updating associative memory was similar between younger and older adults.

Manipulação do contexto da tarefa A-não-B: efeitos no comportamento perseverativo no olhar

Pós-graduação em Ciências da Motricidade - IBRC

Developmental aspects of synaesthesia across the adult lifespan

In synaesthesia, stimuli such as sounds, words or letters trigger experiences of colors, shapes or tastes and the consistency of these experiences is a hallmark of this condition. In this study we investigate for the first time whether there are age-related changes in the consistency of synaesthetic experiences. We tested a sample of more than 400 grapheme-color synaesthetes who have color experiences when they see letters and/or digits with a well-established test of consistency. Our results showed a decline in the number of consistent grapheme-color associations across the adult lifespan. We also assessed age-related changes in the breadth of the color spectrum. The results showed that the appearance of primary colors (i.e., red, blue, and green) was mainly age-invariant. However, there was a decline in the occurrence of lurid colors while brown and achromatic tones occurred more often as concurrents in older age. These shifts in the color spectrum suggest that synaesthesia does not simply fade, but rather undergoes more comprehensive changes. We propose that these changes are the result of a combination of both age-related perceptual and memory processing shifts.

Tracking the Flow of Information through the Hippocampal Formation in the Rat

The hippocampus receives input from upper levels of the association cortex and is implicated in many mnemonic processes, but the exact mechanisms by which it codes and stores information is an unresolved topic. This work examines the flow of information through the hippocampal formation while attempting to determine the computations that each of the hippocampal subfields performs in learning and memory. The formation, storage, and recall of hippocampal-dependent memories theoretically utilize an autoassociative attractor network that functions by implementing two competitive, yet complementary, processes. Pattern separation, hypothesized to occur in the dentate gyrus (DG), refers to the ability to decrease the similarity among incoming information by producing output patterns that overlap less than the inputs. In contrast, pattern completion, hypothesized to occur in the CA3 region, refers to the ability to reproduce a previously stored output pattern from a partial or degraded input pattern. Prior to addressing the functional role of the DG and CA3 subfields, the spatial firing properties of neurons in the dentate gyrus were examined. The principal cell of the dentate gyrus, the granule cell, has spatially selective place fields; however, the behavioral correlates of another excitatory cell, the mossy cell of the dentate polymorphic layer, are unknown. This report shows that putative mossy cells have spatially selective firing that consists of multiple fields similar to previously reported properties of granule cells. Other cells recorded from the DG had single place fields. Compared to cells with multiple fields, cells with single fields fired at a lower rate during sleep, were less likely to burst, and were more likely to be recorded simultaneously with a large population of neurons that were active during sleep and silent during behavior. These data suggest that single-field and multiple-field cells constitute at least two distinct cell classes in the DG. Based on these characteristics, we propose that putative mossy cells tend to fire in multiple, distinct locations in an environment, whereas putative granule cells tend to fire in single locations, similar to place fields of the CA1 and CA3 regions. Experimental evidence supporting the theories of pattern separation and pattern completion comes from both behavioral and electrophysiological tests. These studies specifically focused on the function of each subregion and made implicit assumptions about how environmental manipulations changed the representations encoded by the hippocampal inputs. However, the cell populations that provided these inputs were in most cases not directly examined. We conducted a series of studies to investigate the neural activity in the entorhinal cortex, dentate gyrus, and CA3 in the same experimental conditions, which allowed a direct comparison between the input and output representations. The results show that the dentate gyrus representation changes between the familiar and cue altered environments more than its input representations, whereas the CA3 representation changes less than its input representations. These findings are consistent with longstanding computational models proposing that (1) CA3 is an associative memory system performing pattern completion in order to recall previous memories from partial inputs, and (2) the dentate gyrus performs pattern separation to help store different memories in ways that reduce interference when the memories are subsequently recalled.

Developmental aspects of synaesthesia across the adult lifespan

In synaesthesia, stimuli such as sounds, words or letters trigger experiences of colors, shapes or tastes and the consistency of these experiences is a hallmark of this condition. In this study we investigate for the first time whether there are age-related changes in the consistency of synaesthetic experiences. We tested a sample of more than 400 grapheme-color synaesthetes who have color experiences when they see letters and/or digits with a well-established test of consistency. Our results showed a decline in the number of consistent grapheme-color associations across the adult lifespan. We also assessed age-related changes in the breadth of the color spectrum. The results showed that the appearance of primary colors (i.e., red, blue, and green) was mainly age-invariant. However, there was a decline in the occurrence of lurid colors while brown and achromatic tones occurred more often as concurrents in older age. These shifts in the color spectrum suggest that synaesthesia does not simply fade, but rather undergoes more comprehensive changes. We propose that these changes are the result of a combination of both age-related perceptual and memory processing shifts.

Development of synesthesia across the adult lifespan

In synesthesia, stimuli such as sounds, words or letters trigger experiences of colors, shapes or tastes. The consistency of these experiences is a hallmark of this condition. In a previous study we investigated for the first time whether there are age-related changes in the consistency of synesthetic experiences. Using a cross-sectional approach, we tested a sample of more than 400 grapheme-color synesthetes who have color experiences when they see letters and/or digits with a well-established test of consistency. Our results showed a decline in the number of consistent grapheme-color associations across the adult lifespan. We also assessed age-related changes in the breadth of the color spectrum. The results showed that the appearance of primary colors (i.e., red, blue, and green) was mainly age-invariant. However, there was a decline in the occurrence of lurid colors while brown and achromatic tones occurred more often as concurrents in older age. These shifts in the color spectrum suggest that synesthesia does not simply fade, but rather undergoes more comprehensive changes. These changes may be the result of a combination of both age-related perceptual and memory processing shifts. I will present the results of a second wave of data acquisition after a one-year interval to investigate the longitudinal age-related trajectory of the consistency of synesthetic experiences.

Aportación al análisis de espectros de radiación: estructuras y algoritmos

A lo largo del presente trabajo se investiga la viabilidad de la descomposición automática de espectros de radiación gamma por medio de algoritmos de resolución de sistemas de ecuaciones algebraicas lineales basados en técnicas de pseudoinversión. La determinación de dichos algoritmos ha sido realizada teniendo en cuenta su posible implementación sobre procesadores de propósito específico de baja complejidad. En el primer capítulo se resumen las técnicas para la detección y medida de la radiación gamma que han servido de base para la confección de los espectros tratados en el trabajo. Se reexaminan los conceptos asociados con la naturaleza de la radiación electromagnética, así como los procesos físicos y el tratamiento electrónico que se hallan involucrados en su detección, poniendo de relieve la naturaleza intrínsecamente estadística del proceso de formación del espectro asociado como una clasificación del número de detecciones realizadas en función de la energía supuestamente continua asociada a las mismas. Para ello se aporta una breve descripción de los principales fenómenos de interacción de la radiación con la materia, que condicionan el proceso de detección y formación del espectro. El detector de radiación es considerado el elemento crítico del sistema de medida, puesto que condiciona fuertemente el proceso de detección. Por ello se examinan los principales tipos de detectores, con especial hincapié en los detectores de tipo semiconductor, ya que son los más utilizados en la actualidad. Finalmente, se describen los subsistemas electrónicos fundamentales para el acondicionamiento y pretratamiento de la señal procedente del detector, a la que se le denomina con el término tradicionalmente utilizado de Electrónica Nuclear. En lo que concierne a la espectroscopia, el principal subsistema de interés para el presente trabajo es el analizador multicanal, el cual lleva a cabo el tratamiento cualitativo de la señal, y construye un histograma de intensidad de radiación en el margen de energías al que el detector es sensible. Este vector N-dimensional es lo que generalmente se conoce con el nombre de espectro de radiación. Los distintos radionúclidos que participan en una fuente de radiación no pura dejan su impronta en dicho espectro. En el capítulo segundo se realiza una revisión exhaustiva de los métodos matemáticos en uso hasta el momento ideados para la identificación de los radionúclidos presentes en un espectro compuesto, así como para determinar sus actividades relativas. Uno de ellos es el denominado de regresión lineal múltiple, que se propone como la aproximación más apropiada a los condicionamientos y restricciones del problema: capacidad para tratar con espectros de baja resolución, ausencia del concurso de un operador humano (no supervisión), y posibilidad de ser soportado por algoritmos de baja complejidad capaces de ser instrumentados sobre procesadores dedicados de alta escala de integración. El problema del análisis se plantea formalmente en el tercer capítulo siguiendo las pautas arriba mencionadas y se demuestra que el citado problema admite una solución en la teoría de memorias asociativas lineales. Un operador basado en este tipo de estructuras puede proporcionar la solución al problema de la descomposición espectral deseada. En el mismo contexto, se proponen un par de algoritmos adaptativos complementarios para la construcción del operador, que gozan de unas características aritméticas especialmente apropiadas para su instrumentación sobre procesadores de alta escala de integración. La característica de adaptatividad dota a la memoria asociativa de una gran flexibilidad en lo que se refiere a la incorporación de nueva información en forma progresiva.En el capítulo cuarto se trata con un nuevo problema añadido, de índole altamente compleja. Es el del tratamiento de las deformaciones que introducen en el espectro las derivas instrumentales presentes en el dispositivo detector y en la electrónica de preacondicionamiento. Estas deformaciones invalidan el modelo de regresión lineal utilizado para describir el espectro problema. Se deriva entonces un modelo que incluya las citadas deformaciones como una ampliación de contribuciones en el espectro compuesto, el cual conlleva una ampliación sencilla de la memoria asociativa capaz de tolerar las derivas en la mezcla problema y de llevar a cabo un análisis robusto de contribuciones. El método de ampliación utilizado se basa en la suposición de pequeñas perturbaciones. La práctica en el laboratorio demuestra que, en ocasiones, las derivas instrumentales pueden provocar distorsiones severas en el espectro que no pueden ser tratadas por el modelo anterior. Por ello, en el capítulo quinto se plantea el problema de medidas afectadas por fuertes derivas desde el punto de vista de la teoría de optimización no lineal. Esta reformulación lleva a la introducción de un algoritmo de tipo recursivo inspirado en el de Gauss-Newton que permite introducir el concepto de memoria lineal realimentada. Este operador ofrece una capacidad sensiblemente mejorada para la descomposición de mezclas con fuerte deriva sin la excesiva carga computacional que presentan los algoritmos clásicos de optimización no lineal. El trabajo finaliza con una discusión de los resultados obtenidos en los tres principales niveles de estudio abordados, que se ofrecen en los capítulos tercero, cuarto y quinto, así como con la elevación a definitivas de las principales conclusiones derivadas del estudio y con el desglose de las posibles líneas de continuación del presente trabajo.---ABSTRACT---Through the present research, the feasibility of Automatic Gamma-Radiation Spectral Decomposition by Linear Algebraic Equation-Solving Algorithms using Pseudo-Inverse Techniques is explored. The design of the before mentioned algorithms has been done having into account their possible implementation on Specific-Purpose Processors of Low Complexity. In the first chapter, the techniques for the detection and measurement of gamma radiation employed to construct the spectra being used throughout the research are reviewed. Similarly, the basic concepts related with the nature and properties of the hard electromagnetic radiation are also re-examined, together with the physic and electronic processes involved in the detection of such kind of radiation, with special emphasis in the intrinsic statistical nature of the spectrum build-up process, which is considered as a classification of the number of individual photon-detections as a function of the energy associated to each individual photon. Fbr such, a brief description of the most important matter-energy interaction phenomena conditioning the detection and spectrum formation processes is given. The radiation detector is considered as the most critical element in the measurement system, as this device strongly conditions the detection process. Fbr this reason, the characteristics of the most frequent detectors are re-examined, with special emphasis on those of semiconductor nature, as these are the most frequently employed ones nowadays. Finally, the fundamental electronic subsystems for preaconditioning and treating of the signal delivered by the detector, classically addresed as Nuclear Electronics, is described. As far as Spectroscopy is concerned, the subsystem most interesting for the scope covered by the present research is the so-called Multichannel Analyzer, which is devoted to the cualitative treatment of the signal, building-up a hystogram of radiation intensity in the range of energies in which the detector is sensitive. The resulting N-dimensional vector is generally known with the ñame of Radiation Spectrum. The different radio-nuclides contributing to the spectrum of a composite source will leave their fingerprint in the resulting spectrum. Through the second chapter, an exhaustive review of the mathematical methods devised to the present moment to identify the radio-nuclides present in the composite spectrum and to quantify their relative contributions, is reviewed. One of the more popular ones is the so-known Múltiple Linear Regression, which is proposed as the best suited approach according to the constraints and restrictions present in the formulation of the problem, i.e., the need to treat low-resolution spectra, the absence of control by a human operator (un-supervision), and the possibility of being implemented as low-complexity algorithms amenable of being supported by VLSI Specific Processors. The analysis problem is formally stated through the third chapter, following the hints established in this context, and it is shown that the addressed problem may be satisfactorily solved under the point of view of Linear Associative Memories. An operator based on this kind of structures may provide the solution to the spectral decomposition problem posed. In the same context, a pair of complementary adaptive algorithms useful for the construction of the solving operator are proposed, which share certain special arithmetic characteristics that render them specially suitable for their implementation on VLSI Processors. The adaptive nature of the associative memory provides a high flexibility to this operator, in what refers to the progressive inclusión of new information to the knowledge base. Through the fourth chapter, this fact is treated together with a new problem to be considered, of a high interest but quite complex nature, as is the treatment of the deformations appearing in the spectrum when instrumental drifts in both the detecting device and the pre-acconditioning electronics are to be taken into account. These deformations render the Linear Regression Model proposed almost unuseful to describe the resulting spectrum. A new model including the drifts is derived as an extensión of the individual contributions to the composite spectrum, which implies a simple extensión of the Associative Memory, which renders this suitable to accept the drifts in the composite spectrum, thus producing a robust analysis of contributions. The extensión method is based on the Low-Amplitude Perturbation Hypothesis. Experimental practice shows that in certain cases the instrumental drifts may provoke severe distortions in the resulting spectrum, which can not be treated with the before-mentioned hypothesis. To cover also these less-frequent cases, through the fifth chapter, the problem involving strong drifts is treated under the point of view of Non-Linear Optimization Techniques. This reformulation carries the study to the consideration of recursive algorithms based on the Gauss-Newton methods, which allow the introduction of Feed-Back Memories, computing elements with a sensibly improved capability to decompose spectra affected by strong drifts. The research concludes with a discussion of the results obtained in the three main levéis of study considerad, which are presented in chapters third, fourth and fifth, toghether with the review of the main conclusions derived from the study and the outline of the main research lines opened by the present work.

A neuronal model of a global workspace in effortful cognitive tasks

A minimal hypothesis is proposed concerning the brain processes underlying effortful tasks. It distinguishes two main computational spaces: a unique global workspace composed of distributed and heavily interconnected neurons with long-range axons, and a set of specialized and modular perceptual, motor, memory, evaluative, and attentional processors. Workspace neurons are mobilized in effortful tasks for which the specialized processors do not suffice. They selectively mobilize or suppress, through descending connections, the contribution of specific processor neurons. In the course of task performance, workspace neurons become spontaneously coactivated, forming discrete though variable spatio-temporal patterns subject to modulation by vigilance signals and to selection by reward signals. A computer simulation of the Stroop task shows workspace activation to increase during acquisition of a novel task, effortful execution, and after errors. We outline predictions for spatio-temporal activation patterns during brain imaging, particularly about the contribution of dorsolateral prefrontal cortex and anterior cingulate to the workspace.