Siemens Aktiengesellschaft

Hans-Georg Zimmermann

Christoph Tietz

Ralph Grothmann

Ralf Neuneier

Steffen Udluft

Volker Tresp

Kai Heesche

Anton Maximilian Schäfer

Jochen Cleve

Achim Mueller

Ralph Neuneier

Stefan Siekmann

Computer-assisted analysis of a data record from observations is provided. The data record contains, for each observation, a data vector that includes values of input variables and a value of a target variable. A neuron network structure is learned from differently initialized neuron networks based on the data record. The neuron networks respectively include an input layer, one or more hidden layers, and an output layer. The input layer includes at least a portion of the input variables, and the output layer includes the target variable. The neuron network structure outputs the mean value of the target variables of the output layers of the neuron networks. Sensitivity values are determined by the neuron network structure and stored. Each sensitivity value is assigned an observation and an input variable. The sensitivity value includes the derivative of the target variable of the assigned observation with respect to the assigned input variable.

Computer-assisted analysis of a data record from observations

Disclosed is a method for the computer-assisted modeling of a technical system. One or more output vectors are modeled dependent on one or more input vectors by the learning process of a neural network on the basis of training data of known input vectors and output vectors. Each output vector comprises one or more operating variables of the technical system, and each input vector comprises one or more input variables that influence the operating variable(s). The neural network is a feedforward network with an input layer, a plurality of hidden layers, and an output layer. The output layer comprises a plurality of output clusters, each of which consists of one or more output neurons, the plurality of output clusters corresponding to the plurality of hidden layers. Each output cluster describes the same output vector and is connected to another hidden layer.

Method for the computer-assisted modeling of a wind power installation or a photovoltaic installation with a feed forward neural network

A method for the computer-aided learning of a recurrent neural network for modeling a dynamic system which is characterized at respective times by an observable vector with one or more observables as entries is provided. The neural network includes both a causal network with a flow of information that is directed forwards in time and a retro-causal network with a flow of information which is directed backwards in time. The states of the dynamic system are characterized by first state vectors in the causal network and by second state vectors in the retro-causal network, wherein the state vectors each contain observables for the dynamic system and also hidden states of the dynamic system. Both networks are linked to one another by a combination of the observables from the relevant first and second state vectors and are learned on the basis of training date including known observables vectors.

Method for the computer-aided learning of a recurrent neural network for modeling a dynamic system

A method for the computer-assisted control and/or regulation of a technical system is provided. The method includes two steps, namely modeling the dynamic behavior of the technical system with a recurrent neural network using training data, the recurrent neural network includes states and actions determined using a simulation model at different times and learning an action selection rule by the recurrent neural network to a further neural network. The method can be used with any technical system in order to control the system in an optimum computer-assisted manner. For example, the method can be used in the control of a gas turbine.

Method for the computer-assisted control and/or regulation of a technical system where the dynamic behavior of the technical system is modeled using a recurrent neural network

An approximation is determined for the future system behavior by a similarity analysis using a previously known behavior of the dynamic system, whereupon the future system behavior is determined by using the approximation for the future behavior of the dynamic system as well as a neuronal network structure, especially a causal retro-causal network (causality analysis).

Method, computer program with program code means, and computer program product for determining a future behavior of a dynamic system

The invention relates to a method for speech processing in which input variables containing speech features are mapped onto output variables. In the mapping process, the input variables are weighted and/or identical maps are produced for different sets of input variables and at least one output variable.

Methods for speech processing

Computation elements are connected to one another with a first subsystem having a first input computation element, to which time series values, which each describe one state of a system in a first state space at a time, can be supplied. The first input computation element is connected to a first intermediate computation element, by which a state of the system can be described in a second state space at a time. In a second subsystem a second intermediate computation element, by which a state of the system can be described in the second state space at a time, is connected to a first output computation element, on which a first output signal can be tapped off. In a third subsystem a third intermediate computation element, by which a state of the system can be described in the second state space at a time, is connected to a second output computation element, on which a second output signal can be tapped off. The first subsystem, the second subsystem and the third subsystem are each connected to one another by a coupling between the intermediate computation elements. Weights, which are each associated with one connection between two intermediate computation elements are equal to one another, and weights which are each associated with a connection between an intermediate computation element and an output computation element are equal to one another.

System and method for training and using interconnected computation elements to determine a dynamic response on which a dynamic process is based

An input signal is transformed into a predetermined space. Transformation computer elements are connected to one another such that transformed signals can be taken at the transformation computer elements, whereby at least three transformed signals relate to respectively successive points in time. Composite computer elements are respectively connected to two transformation computer elements. Further, a first output computer element is provided at which an output signal that describes a system status at a point in time can be taken. The first output computer element is connected to the transformation computer elements. Further, a second output computer element is provided that is connected to the composite computer elements and given whose employment a predetermined condition can be taken into consideration when training the arrangement.

ASSEMBLY OF INTERCONNECTED COMPUTING ELEMENTS, METHOD FOR COMPUTER-ASSISTED DETERMINATION OF A DYNAMICS WHICH IS THE BASE OF A DYNAMIC PROCESS, AND METHOD FOR COMPUTER-ASSISTED TRAINING OF AN ASSEMBLY OF INTERCONNECTED ELEMENTS

A set of fuzzy rules (FR) is mapped onto a neural network (NN) (,). The neural network (NN) is trained (,), and weights (w,) and/or neurons (NE) of the neural network (NN) are pruned or grown (,). A new neural network (NNN) formed in this way is mapped onto a new fuzzy rule set (NFR) (,).

Method for optimizing a set of fuzzy rules using a computer

The training phase of a neural network NN is stopped before an error function, which is to be minimized in the training phase, reaches a minimum (,). A first variable (EG) is defined using, for example, the optimal brain damage method or the optimal brain surgeon method, on the assumption that the error function is at the minimum. Furthermore, a second variable (ZG) is determined which provides an indication of the manner in which the value of the error function varies when a weight (w,) is removed from the neural network (NN). The first variable (EG) and the second variable (ZG) are used to classify the weight (w,) as being suitable or unsuitable for removal from the neural network (NN).

Method for determination of weights, suitable for elimination, of a neural network using a computer

A method and apparatus for computer-supported generation of at least one artificial training data vector for a neural network is provided wherein a residual error is determined after a training of a neural network has occurred. A backward error is then determined from the residual error. Artificial training data vectors are generated from a statistical random process that is based on a statistical distribution, such that the respective backward error for an input of the neural network is taken into consideration.

Growing community of inventors

Starnberg/Percha, Germany

Hans-Georg Zimmermann