Computer Associates Think, Inc.

Ca, Inc.

Ai Ware, Inc.

Zhuo Meng

Yoh-Han Pao

Baofu Duan

Ronald J Cass

David Eugene Huddleston

Qian Yang

Xinyu Mao

A method of training a neural net includes receiving a plurality of sets of data, each set representative of a plurality of inputs to the neural net and a resulting at least one output from the neural net and calculating a plurality of network weights for the neural network based on the received plurality of sets of data. Calculating the plurality of network weights including attributing greater weight in the calculation to at least one set of the plurality of sets of data than at least one other set of the plurality of sets of data.

Weighted pattern learning for neural networks

A method to add expression level fuzzy rule processing capability to existing crisp rule engine without requiring extensive modifications to such rule engine. A system and computer implemented method receive a data set at a crisp rule engine. If the data set involves a fuzzy rule, a callback to a fuzzy rule broker is done. The fuzzy rule broker selects one of multiple tailored processing modules for processing the fuzzy rule. The result of processing the fuzzy rule may be converted to a crisp result and returned to the crisp rule engine.

Fuzzy rule handling

A method for hierarchical visualization of multi-dimensional data is provided. A first dimension-reduction process is applied to a multi-dimensional data set to obtain a first visualization. A subset of the multi-dimensional data set associated with a selected region of the dimension-reduced first visualization is selected. A second dimension-reduction process is applied to the selected subset of the multi-dimensional data set to obtain at least one additional visualization.

Viewing multi-dimensional data through hierarchical visualization

An apparatus and method for processing mixed data for a selected task is provided. An input transformation module converts mixed data into converted data. A functional mapping module processes the converted data to provide a functional output for the selected task. The selected task may be one or a combination of a variety of possible tasks, including search, recall, prediction, classification, etc. For example, the selected task may be for data mining, database search, targeted marketing, computer virus detection, etc.

Processing mixed numeric and symbolic data encodings using scaling at one distance of at least one dimension, clustering, and a signpost transformation

A method for automated feature selection is provided. One or more initial sets of features are generated and evaluated to determine quality scores for the feature sets. Selected ones of the feature sets are (i) chosen according to the quality scores and modified to generate a generation of modified feature sets, (ii) the modified feature sets are evaluated to determine quality scores for the modified feature sets, and (i) and (ii) are repeated until a modified feature set is satisfactory.

Method and apparatus for automated feature selection

Method of incrementally forming and adaptively updating a neural net model are provided. A function approximation node is incrementally added to the neural net model. Function parameters for the function approximation node are determined and function parameters of other nodes in the neural network model are updated, by using the function parameters of the other nodes prior to addition of the function approximation node to the neural network model.

Automatic neural-net model generation and maintenance

A model maintenance method is provided. If accuracy of prediction by a current model through consultation with new data is determined to be below a predetermined threshold, a compound model is formed by supplementing the current model with a local net trained with the new data.

Automatic model maintenance through local nets

Methods for feature selection based on hierarchical local-region analysis of feature relationships in a data set are provided.

Hierarchical determination of feature relevancy

A method for feature selection based on hierarchical local-region analysis of feature characteristics in a data set of mixed data type is provided. A data space associated with a mixed-type data set is partitioned into a hierarchy of plural local regions. A relationship metric (for example, a similarity correlation metric) is used to evaluate for each local region a relationship measure between input features and a target. One or more relevant features is identified, by using the relationship measure for each local region.

Hierarchical determination of feature relevancy for mixed data types

An adaptive learning method for automated maintenance of a neural net model is provided. The neural net model is trained with an initial set of training data. Partial products of the trained model are stored. When new training data are available, the trained model is updated by using the stored partial products and the new training data to compute weights for the updated model.

Adaptive learning enhancement to automated model maintenance

The subject system provides reduced-dimension mapping of pattern data. Mapping is applied through conventional single-hidden-layer feed-forward neural network with non-linear neurons. According to one aspect of the present invention, the system functions to equalize and orthogonalize lower dimensional output signals by reducing the covariance matrix of the output signals to the form of a diagonal matrix or constant times the identity matrix. The present invention allows for visualization of large bodies of complex multidimensional data in a relatively 'topologically correct' low-dimension approximation, to reduce randomness associated with other methods of similar purposes, and to keep the mapping computationally efficient at the same time.

Visualization and self-organization of multidimensional data through equalized orthogonal mapping

The subject system provides reduced-dimension mapping of pattern data. Mapping is applied through conventional single-hidden-layer feed-forward neural network with non-linear neurons. According to one aspect of the present invention, the system functions to equalize and orthogonalize lower dimensional output signals by reducing the covariance matrix of the output signals to the form of a diagonal matrix or constant times the identity matrix. The present invention allows for visualization of large bodies of complex multidimensional data in a relatively &ldquo;topologically correct&rdquo; low-dimension approximation, to reduce randomness associated with other methods of similar purposes, and to keep the mapping computationally efficient at the same time.

Growing community of inventors

Broadview Heights, OH, United States of America

Zhuo Meng