Siemens Aktiengesellschaft

Max-planck-gesellschaft Zur F&#xf6;rderung Der Wissenschaften E. V.

Seimens Aktiengesellschaft

Gustavo Deco

Martin Stetter

Christian Schittenkopf

Bernd Schuermann

Norbert Galm

Bernd Schurmann

Silvia Corchs

Louis-J Dubé

Bernd Sch&uuml;rmann

Bernd Schürmann

Laura Martignon

Miruna Szabo

Jan Storck

There is described a method for computer-aided learning of a neural network, with a plurality of neurons in which the neurons of the neural network are divided into at least two layers, comprising a first layer and a second layer crosslinked with the first layer. In the first layer input information is respectively represented by one or more characteristic values from one or several characteristics, wherein every characteristic value comprises one or more neurons of the first layer. A plurality of categories is stored in the second layer, wherein every category comprises one or more neurons of the second layer. For one or several pieces of input information, respectively at least one category in the second layer is assigned to the characteristic values of the input information in the first layer. Input information is entered into the first layer and subsequently at least one state variable of the neural network is determined and compared to the at least one category of this input information assigned in a preceding step. The crosslinking between the first and second layer is changed depending on the comparison result from a preceding step.

Method for computer-aided learning of a neural network and neural network

In a method, arrangement and computer program for analysis of neuronal activities in neuronal areas, signals are recorded, with each signal describing the neuronal activity in one of the neuronal areas. A matchable coupling forms the basis of all signals, described by the use of matchable coupling parameters that describe the statistical relationship between the signals. Probabilities for an occurrence of the signals are determined, whereby a statistical distribution is the basis of the signals. The matchable coupling parameters are determined by optimization of the probabilities, hence matched and analyzed.

Method and arrangement and computer programme with programme code means and computer programme products for the analysis of neuronal activities in neuronal areas

A method for evaluating an image (fMRI-image) of the brain that has been obtained by functional magnetic resonance tomography is provided. According to the method, a neuronal network is used to simulate the activities of the brain. Supposed disorders of the brain are simulated in the neuronal network (as a disturbed neuronal network). The activities determined in the brain can be artificially simulated in the model and its effect on the complex synergy of the areas of the brain can be quantified. The comparison with the fMRI image or fMRI activity pattern relating to the patient enables the cause of the disorders to be localized, thus leading to a successful diagnosis.

Evaluation of images of the brain obtained by means of functional magnetic resonance tomography

In a pattern recognition system, a pattern which is to be recognized later is prescribed in a learning phase. This pattern is detected sequentially, that is to say the informative areas of the pattern are detected and, moreover, the spatial relationship between the areas is also stored. In the recognition phase, a hypothesis which indicates a presumed pattern and, furthermore, indicates where such further prominent areas should be located in the pattern to be recognized if the presumption is correct, is generated on the basis of the acquired data of a first area of a pattern to be recognized, and on the basis of the stored data. Thus, patterns are learned through their location information, on the one hand, and through their spatial relationship to each other, on the other hand, stored and then re-recognized. The application options of the present invention reside, for example, in robotics, text analysis and image analysis, but also in the field of medical technology (for example in automatic tumor detection in mammographies, etc).

Pattern recognition

An arrangement and method are presented that enable a prediction of an abnormality and implement a suitable action opposing the abnormality. An information flow underlying a dynamic system is interpreted and a prediction quantity that comprises the abnormality as characterizing quantity of the dynamic system is determined from it. A neural network is trained with measured data of the system. After the training, the abnormality can be indicated on the basis of the prediction quantity before it occurs and the occurrence can be opposed with suitable measures.

Arrangement for predicting an abnormality of a system and for carrying out an action which counteracts the abnormality

The present invention relates to a system and a method for recognizing a prescribed object, having a storage device for storing attribute information relating to the prescribed object, detecting means for detecting attributes in a detection range and for outputting corresponding detection information, first processing means for processing, in parallel and separately for each possible attribute type, the detection information for the detection means by using the attribute information from the storage device and for outputting corresponding processing information, and second processing means for processing the process information from the first processing means and for outputting the information for determining the position of the prescribed object in the detection range.

A time series that is established by a measured signal of a dynamic system, for example a quotation curve on the stock market, is modelled according to its probability density in order to be able to make a prediction of future values. A non-linear Markov process of the order m is suited for describing the conditioned probability densities. A neural network is trained according to the probabilities of the Markov process using the maximum likelihood principle, which is a training rule for maximizing the product of probabilities. The neural network predicts a value in the future for a prescribable number of values m from the past of the signal to be predicted. A number of steps in the future can be predicted by iteration. The order m of the non-linear Markov process, which corresponds to the number of values from the past that are important in the modelling of the conditioned probability densities, serves as parameter for improving the probability of the prediction.

Method of classifying statistical dependency of a measurable series of statistical values

A method for determining conditioned entropies for a prescribable plurality of future sampling times for a set of samples based upon an information flow. A classification of a time series is implemented on the basis of the information flow. The information flow reflects nonlinear correlations between the samples. A classification is thus possible between those time series whose samples are non-linearly correlated and those time series whose samples are stochastically independant.

Method conducted in a computer for classification of a time series having a prescribable number of samples

A method for classifying a time series, that includes a prescribable plurality of samples, with a computer wherein generalized correlation integral is determined for at least a part of samples of a time series. A functions family of an entropy function is determined from the values of the generalized correlation integral. A plurality of considered future samples is thereby employed as a family parameter of the functions family. The time series is classified into various types of characteristic processes from the curve of the functions family of the entropy function.

Growing community of inventors

Neubiberg, Germany

Gustavo Deco