Péter Érdi

Prediction of emerging technologies based on analysis of the US patent citation network

The network of patents connected by citations is an evolving graph, which provides a representation of the innovation process. A patent citing another implies that the cited patent reflects a piece of previously existing knowledge that the citing patent builds upon. A methodology presented here (1) identifies actual clusters of patents: i.e., technological branches, and (2) gives predictions about the temporal changes of the structure of the clusters. A predictor, called the citation vector, is defined for characterizing technological development to show how a patent cited by other patents belongs to various industrial fields. The clustering technique adopted is able to detect the new emerging recombinations, and predicts emerging new technology clusters. The predictive ability of our new method is illustrated on the example of USPTO subcategory 11, Agriculture, Food, Textiles. A cluster of patents is determined based on citation data up to 1991, which shows significant overlap of the class 442 formed at the beginning of 1997. These new tools of predictive analytics could support policy decision making processes in science and technology, and help formulate recommendations for action.

The KIV model—nonlinear spatio-temporal dynamics of the primordial vertebrate forebrain

EEG measurements indicate the presence of common-mode, coherent oscillations in various cortical areas. In previous studies the KIII model has been introduced, which interprets the experimental observation as nonlinear, spatially distributed dynamical oscillations of coupled neural populations. In this paper we combine multiple KIII sets into the KIV model, which approximates the operation of the basic vertebrate forebrain together with the basal ganglia and motor systems. This paper outlines a summary description of the essential components of the KIV model, as the basis for future modeling of their cooperative dynamics guided by analysis of

Artificial Neural Networks and Machine Learning – ICANN 2012

A complex-valued multilayer perceptron (MLP) can approximate a periodic or unbounded function, which cannot be easily realized by a real-valued MLP. Its search space is full of crevasse-like forms having huge condition numbers; thus, it is very hard for existing methods to perform efficient search in such a space. The space also includes the structure of reducibility mapping. The paper proposes a new search method for a complex-valued MLP, which employs both eigen vector descent and reducibility mapping, aiming to stably find excellent solutions in such a space. Our experiments showed the proposed method worked well.

Chaos and learning in the olfactory bulb

A mathematical model is given for describing activity dynamics, learning, and associative memory in the olfactory bulb. Numerical bifurcation analysis and the calculation of Lyapunov‐exponents suggest that chaotic behavior only occurs in the case of strong excitatory coupling in the mitral layer. A Hebbian‐type learning rule, supplemented with a nonlinear decay term and a selective decreasing term, is defined and analyzed. Slow learning modifies the bulbar activity dynamics hence it plays a crucial role in odor information processing.

Modulation of septo-hippocampal θ activity by GABAA receptors: an experimental and computational approach

Theta frequency oscillation of the septo-hippocampal system has been considered as a prominent activity associated with cognitive function and affective processes. It is well documented that anxiolytic drugs diminish septo-hippocampal oscillatory Theta activity contributing to their either therapeutic or unwanted side effects. In the present experiments we applied a combination of computational and physiological techniques to explore the functional role of GABAA receptors in Theta oscillation. In electrophysiological experiments extracellular single unit recordings were performed from medial septum/diagonal band of Broca with simultaneous hippocampal (CA1) electroencephalogram (EEG) recordings from anesthetized rats. Neurotransmission at GABAA receptors were modulated by means of pharmacological tools: the actions of the GABAA receptor positive allosteric modulator diazepam and inverse agonist/negative allosteric modulator FG-7142 were evaluated on septo-hippocampal activity. Systemic administration of diazepam inhibited, whereas FG-7142 enhanced Theta oscillation of septal neurons and hippocampal EEG Theta activity. In parallel to these experimental observations, a computational model has been constructed by implementing a septal GABA neuron model with a CA1 hippocampal model containing three types of neurons (including oriens and basket interneurons and pyramidal cells; latter modeled by multicompartmental techniques; for detailed model description with network parameters see online addendum: http://geza.kzoo.edu/theta). This connectivity made the network capable of simulating the responses of the septo-hippocampal circuitry to the modulation of GABAA transmission, and the presently described computational model proved suitable to reveal several aspects of pharmacological modulation of GABAA receptors. In addition, computational findings indicated different roles of distinctively located GABAA receptors in theta generation.

Challenges in complex systems science

Abstract FuturICT foundations are social science, complex systems science, and ICT. The main concerns and challenges in the science of complex systems in the context of FuturICT are laid out in this paper with special emphasis on the Complex Systems route to Social Sciences. This include complex systems having: many heterogeneous interacting parts; multiple scales; complicated transition laws; unexpected or unpredicted emergence; sensitive dependence on initial conditions; path-dependent dynamics; networked hierarchical connectivities; interaction of autonomous agents; self-organisation; non-equilibrium dynamics; combinatorial explosion; adaptivity to changing environments; co-evolving subsystems; ill-defined boundaries; and multilevel dynamics. In this context, science is seen as the process of abstracting the dynamics of systems from data. This presents many challenges including: data gathering by large-scale experiment, participatory sensing and social computation, managing huge distributed dynamic and heterogeneous databases; moving from data to dynamical models, going beyond correlations to cause-effect relationships, understanding the relationship between simple and comprehensive models with appropriate choices of variables, ensemble modeling and data assimilation, modeling systems of systems of systems with many levels between micro and macro; and formulating new approaches to prediction, forecasting, and risk, especially in systems that can reflect on and change their behaviour in response to predictions, and systems whose apparently predictable behaviour is disrupted by apparently unpredictable rare or extreme events. These challenges are part of the FuturICT agenda. Graphical abstract

Computational theories on the function of theta oscillations

Neural rhythms can be studied in terms of conditions for their generation, or in terms of their functional significance. The theta oscillation is a particularly prominent rhythm, reported to be present in many brain areas, and related to many important cognitive processes. The generating mechanisms of theta have extensively been studied and reviewed elsewhere; here we discuss ideas that have accumulated over the past decades on the computational roles it may subserve. Theories propose different aspects of theta oscillations as being relevant for their cognitive functions: limit cycle oscillations in neuronal firing rates, subthreshold membrane potential oscillations, periodic modulation of synaptic transmission and plasticity, and phase precession of hippocampal place cells. The relevant experimental data is briefly summarized in the light of these theories. Specific models proposing a function for theta in pattern recognition, memory, sequence learning and navigation are reviewed critically. Difficulties with testing and comparing alternative models are discussed, along with potentially important future research directions in the field.

Self-organizing mechanism for the formation of ordered neural mappings

A model for the formation of ordered neural mappings in general, and of retinotectal connections, in particular is given. The main point came from the theory of “noise induced transitions”, i.e. order may be the result of the interplay between deterministic and random interactions. An activity-dependent self-organizing mechanism is presented in terms of modifiable synapses. Simulation experiments were done not only for the normal ontogenetic development but also for the plastic behaviour of the retinotopic connections.

Dynamics of the olfactory bulb: bifurcations, learning, and memory

A mathematical model for describing dynamic phenomena in the olfactory bulb is presented. The nature of attractors and the bifurcation sequences in terms of the lateral connection strength in the mitral layer are studied numerically. Chaotic activity has only been found in the case of strong excitatory coupling. Synaptic modification-induced transition from oscillation to chaos is demonstrated. A model for a simple associative memory is also presented.

Model-based source localization of extracellular action potentials.

A new model-based analysis method was set up for revealing information encrypted in extracellular spatial potential patterns of neocortical action potentials. Spikes were measured by extracellular linear multiple microelectrode in vivo cats primary auditory cortex and were analyzed based on current source density (CSD) distribution models. Validity of the monopole and other point source approximations were tested on the measured potential patterns by numerical fitting. We have found, that point source models could not provide accurate description of the measured patterns. We introduced a new model of the CSD distribution on a spiking cell, called counter-current model (CCM). This new model was shown to provide better description of the spatial current distribution of the cell during the initial negative deflection of the extracellular action potential, from the onset of the spike to the negative peak. The new model was tested on simulated extracellular potentials. We proved numerically, that all the parameters of the model could be determined accurately based on measurements. Thus, fitting of the CCM allowed extraction of these parameters from the measurements. Due to model fitting, CSD could be calculated with much higher accuracy as done with the traditional method because distance dependence of the spatial potential patterns was explicitly taken into consideration in our method. Average CSD distribution of the neocortical action potentials was calculated and spatial decay constant of the dendritic trees was determined by applying our new method.