Prof. Dr. Phil. Dr. Rer. Nat. Habil. Gustavo Deco is Research Professor at the Institucio Catalana de Recerca i Estudis Avanats and Full Professor (Catedratico) the Pompeu Fabra University (Barcelona) where he is head of the Computational Neuroscience Group and Director of the Center of Brain and Cognition.
He studied Physics at the National University of Rosario (Argentina) where he received his diploma degree in Theoretical Atomic Physics. In 1987, he received his Ph.D. degree in Physics for his thesis on Relativistic Atomic Collisions. In 1987, he was a post doctoral fellow at the University of Bordeaux in France. In the period from 1988 to 1990, he obtained a post doctoral position of the Alexander von Humboldt Foundation at the University of Giessen in Germany. From 1990 to 2003, he has been with the Neural Computing Section at the Siemens Corporate Research Center in Munich, Germany, where he led the Computational Neuroscience Group.
In 1997, he obtained his habilitation (maximal academical degree in Germany) in Computer Science (Dr. Rer. Nat. Habil.) at the Technical University of Munich for his thesis on Neural Learning. In 2001, he received his PhD in Psychology (Dr. phil.) for his thesis on Visual Attention at the Ludwig-Maximilian-University of Munich.
He was lecturer at the universities of Rosario, Frankfurt and Munich. Since 1998 he is Associate Professor at the Technical University of Munich and Honorary Professor at the University of Rosario, and since 2001 Invited Lecturer at the Ludwig-Maximilian- University of Munich. Since 2001 he is also McDonnell-Pew Visiting Fellow of the Centre for Cognitive Neuroscience at the University of Oxford.In 2001 he was awarded the international price of Siemens "Inventor of the Year" for his contribution in statistical learning, models of visual perception, and fMRI based diagnosis of neuropsychiatric diseases.
His research interests include computational neuroscience, neuropsychology, psycholinguistics, biological networks, statistical formulation of neural networks, and chaos theory.
He has published 4 books, more than 170 papers in International Journals, 260 papers in International Conferences and 30 book chapters. He has also 52 patents in Europe, USA, Canada and Japan. Recently, he was awarded with the “Advanced ERC” grant.
Neurodynamical Modeling of Human Cognition
The quickly growing complexity of the modern information society confronts us with the task to select valuable and reasonable information from a bigger and bigger avalanche of complex and highly diverse data, as becomes evident from imaging processing in multimedia applications, human machine interface, surveillance, mobile communication and the internet, just to mention a few examples. On the one hand the rapid development and ubiquity of modern information technology provides us with an increasing quantity and complexity of data, but on the other hand the burden of selecting and evaluating valuable information in an adapted and reasonable way is to date still put on the user. One major reason for this imbalance relates to the fact that the selection of useful information represents an intrinsically difficult task: it requires the ability of a system to adaptively reduce the complexity of the data, to generate meaning from it and to flexibly assess the putative value of this meaning given the present and past status of the environment. In other words, it must be capable of active and intelligent perception, reasoning, planning and decision making: it must have human cognitive abilities.
Our research aims to derive a quantitative understanding of higher human precognitive (perceptual) and cognitive abilities and thereby builds up the technological basis for a quantum leap in the fields of intelligent human-like algorithms and human machine interaction (e.g. multimedia systems). We follow as closely as possible and reasonable the biological counterpart of the human brain.
A quantitative approach can be obtained by developing mathematical neurodynamical models of brain function which are based on the techniques of computational and integrative neuroscience. The neurodynamical approach models the mutual interaction of multiple hierarchical brain areas and include biological details from the levels of synaptic and neural spiking dynamical mechanisms up to the level of global brain activation and behavior. By this, we are able to describe neuronal brain activity both at the local and global level. Our models can and will be constrained by comparing quantitative results and predictions with experiments from various sources and at various levels including neuroanatomy (structural information), cellular electrophysiology (microscopic level), functional brain imaging (mesoscopic level) and psychophysics (macroscopic behavioural level). The requirement to simultaneously explain results generated from experiments of different designs, which address different aspects of human cognition and produce data at different neuroscientific levels will ensure both sufficiently strong constrains of the models and their proximity to the biological counterpart, the human brain. Biologically plausible neurodynamical modeling of cognitive phenomena will be referred to as neurocognitive modeling.
A natural focus for a concrete neurocognitive modeling should be as simple as possible, should involve various aspects of cognition, which enable human beings to deal with a complex sensorial environment (e.g. audiovisual environments), and should be carried by a sufficiently strong body of experimental evidence. Based on these considerations, we investigate mechanisms of human visual cognition, which provide us with the ability of intelligent and flexible analysis of complex visual scenes: We build up a model of neuronal mechanisms of visual attention and how attention is controlled by an interplay of conflict detection/ management and short term memory. This approach follows the hypothesis that cognitive phenomena like visual information selection, reasoning and decision making about which information to select, are generated by the mutual recurrent influence and interaction between brain areas related to visual perception, conflict detection, learning and memory. We are also applying computational neuroscience based techniques in other cognitive areas, like psycholinguistics.
The project fits perfectly in the framework of a future technology for the modern information society: It will provide the basis for human-like intelligent systems which are able to process complex audiovisual data, to flexibly assess their contents in light of memorized and present context and to autonomously draw or at least suggest decisions about further proceeding.
The construction of neurocognitive models requires the concentration of leading edge competencies over a wide range of diverse disciplines: neurobiology, neurophysiology, neuropsychology statistics, computational neuroscience and industrial R&D just to mention major streams. My cooperation partners in Europe satisfy such requirements.
As mentioned above, the modern information society on the one hand - faces an urgent need of new technologies with human-like intelligent capabilities, which will open the gate towards the automatic categorization, interpretation and presentation of information. On the other hand, the rapid development of functional brain imaging techniques such as functional magnetic resonance imaging (fMRI) provide by now the basis for a new approach towards the understanding of human brain function including higher cognitive phenomena. During the past ten years, a strongly growing body of experimental fMRI investigations has been generated, however these studies are generally of descriptive but not of explaining nature functional brain imaging can indirectly measure global brain activation in presence of cognitive tasks and thereby has the potential to bridge the gap between the microscopic level of electrophysiological measurements on animals and the macroscopic level of psychophysical tests, a gap which has been prohibitive to quantitative and biologically plausible modelling of cognition so far, because models could not be sufficiently constrained by experimental findings. Consequently, existing models of cognitive phenomena reflect rather considerations from an engineering point of view, and generally come not even close to human cognitive abilities.
Therefore it is now the right time to make use of the emerging experimental techniques and to attack the task of revealing and quantifying the neuronal mechanisms by which we successfully navigate and operate in our open and complex environment, in other words to construct artificial systems with human-like intelligent abilities by mirroring the biological counterpart instead of following mainly engineering considerations.
The approach of combining psychophysics, neurophysiology and statistical theory to a quantitative neurocognitive modelling project is entirely new and revolutionary, and therefore is timely but at the same time feasible because of the by now established portfolio of experimental and theoretical techniques. This interdisciplinarity is essential for the analysis, design and creation of new techniques and way of communications in audiovisual environments. In other words, perception and the neural substrate of human brain functions will contribute crucially to the systematic development of innovative human machine interfaces and interactions.
Generation of applied research and by-products for other disciplines: A quantitative mathematical description of human brain function will dramatically bring forward the quality and throughput of healthcare for brain diseases, some of which (e.g., Alzheimers disease) belong to the most frequent diseases of the European civilization. Neurocognitive modeling will yield important tools to identify brain diseases and to help identifying their neuronal origins, for example by modelling the brain activations seen in patients suffering from cerebral disorders. Similarly, other aspects including rehabilitation monitoring, therapy planning and drug evaluation will benefit drastically from the ability of quantitative assessment of brain states.
My investigations require research contributions at three levels (experimental measurements statistical data modeling and data fusion, and neurodynamical modeling) followed and paralleled by technological development and the generation of spin-offs in the framework of industrial exploitation. For example, in order to constrain models, electrophysiological (animal model), brain imaging and psychophysical experiments will be carried out while subjects or animals perform suitable cognitive tasks. These tasks, in turn, will be designed in a coherent way by psychologists, imagers and modelers, in order to maximize the value of the experiments for constraining the models while ensuring feasibility under the different experimental conditions. The generated data will be statistically analysed and fused to form the basis of neurodynamical models, the results and predictions of which will be cross-checked against new experiments, and must be steadily exploited to yield new spin-offs. (See my recent book: "Computational Neuroscience of Vision", Oxford University Press, for a more detailed explanation and previous results obtained in this framework). I have been following this approach based on computational and integrative neuroscience since years, and I established for that purpose an effective network of cooperation partners at different laboratories in universities, academia and industry in Europe and USA (e.g. Univ. Oxford, Univ. Paris, Univ. Carnegie-Mellon, Univ. Birmingham, Univ. Munich, Univ. Luebeck, Forschungszentrum Juelich, Univ. Plymouth, Univ. of Bern, NIH Washington, Siemens). The integration of these efforts will then collect the critical mass required to achieve the quantum leap towards a quantitative understanding of human thinking and brain-like technologies.
My Wife and Kids
Links to my books
- 1 of 14
03.83-03.88 Teaching Assistent at the National University of Rosario.
- Theoretical Mechanic (1983-1984)
- Theoretical Electrodynamic (1985-1986)
- Quantum mechanic (1987-1988)
04.1994 Invited Professor at the National University of Rosario.
- Introduction to the Theory of Neural Networks.
1994-1995 Teaching Assistent at the University of Frankfurt
- Introduction to the Theory of Neural Networks.
Since 1994 Privat-Dozent at the Technical University of Munich
- Information Theory and Neural Network.
- Computational Neuroscience.
05.1997 Invited Professor at the National University of Rosario.
- Theoretical Computer Science
Since 01.1999 Honorar Professor at the National University of Rosario.
Since 03.2003 Professor at the University Pompeu Fabra.
- Taller de Modelización y Simulación I
- Taller de Modelización y Simulación II
- Introduction to Computational Neuroscience (PhD Course)