International Business Machines Corporation

Sony Corporation

Globalfoundries Inc.

Tsuyoshi Ide

Tetsuro Morimura

Hisashi Kashima

Keisuke Inoue

Bin Tong

Jayant R Kalagnanam

Shoko Suzuki

Hiroaki Etoh

Ryo Hirade

Yoichi Taira

Toshihiro Takahashi

Hiroki Yanagisawa

Hidetoshi Numata

Hironori Takeuchi

Toshiyuki Yamane

Akira Tajima

Yuta Tsuboi

Takashi Imamichi

Raymond Harry Rudy

Dzung Tien Phan

Shohei Hido

Akiko Nishikai

Kunikazu Yoda

Hideyuki Mizuta

Rikiya Takahashi

Masato Nakamura

Koji Ashizaki

Jiri Navratil

Dung Tien Phan

Amit Dhurandhar

Hiroshi Udagawa

Yuji Kawamura

Naoki Abe

Yushi Ihara

Moninder Singh

Spyridon Papadimitriou

Kohei Nojiri

Takahiro Nishikawa

Kensuke Baba

Ichiro Ujiie

Chizuru Makita

Toshiya Izumi

Matthew Menickelly

A computer-implemented method, a computer program product, and a computer system for diagnosing anomalies detected by a black-box machine learning model. A computer determines a local variance of a test sample in a test dataset, where the local variance represents uncertainty of a prediction by the black-box machine learning model. The computer initializes optimal compensations for the test sample, where the optimal compensations are optimal perturbations to test sample values of respective components of a multivariate input variable. The computer determines local gradients for the test sample. Based on the local variance and the local gradients, the computer updates the optimal compensations until convergences of the optimal compensations are reached. Using the optimal compensations, the computer diagnoses the anomalies detected by the black-box machine learning model.

Diagnosing anomalies detected by black-box machine learning models

A first dependency graph is constructed based on a first data set by solving an objective function constrained with a maximum number of non-zeros and formulated with a regularization term comprising a quadratic penalty to control sparsity. The quadratic penalty in constructing the second dependency graph is determined as a function of the first data set. A second dependency graph is constructed based on a second data set by solving the objective function constrained with the maximum number of non-zeros and formulated with the regularization term comprising a quadratic penalty. The quadratic penalty in constructing the second dependency graph is determined as a function of the first data set and the second data set. An anomaly score is determined for each of a plurality of sensors based on comparing the first dependency graph and the second dependency graph, nodes of which represent sensors.

Learning sparsity-constrained gaussian graphical models in anomaly detection

A probability distribution of a manufacturing system's performance conditioned on a training dataset comprising a historical tensor and associated performance metric of a reference period is learned. An input tensor associated with a time window and the input tensor's associated performance metric may be received. The input tensor includes at least multiple sensor variables associated with the manufacturing system and multiple steps of the manufacturing system's manufacturing process. Based on the probability distribution, an overall change is determined between the training dataset's relationship of the historical tensor and associated performance metric, and the relationship of the input tensor and the input tensor's associated performance metric. Based on the probability distribution, contribution of at least one of the multiple variables and the multiple steps to the overall change is determined. An action is automatically triggered in the manufacturing system which reduces the overall change.

Incorporating change diagnosis using probabilistic tensor regression model for improving processing of materials

A system and method for maintaining health of a fleet of assets implementing an asset maintenance framework for collective anomaly detection that provides for a more accurate maintenance planning solution for the fleet or assets that may be prioritized. Based on a Bayesian multi-task multi-modal sparse mixture of sparse Gaussian graphical models (MTL-MM GGM), the methods combine the variational Bayes framework with (1) Laplace prior-based sparse structure learning and (2) an-based sparse mixture weight selection approach. Dual sparsity is guaranteed over both variable-variable dependency and mixture components to efficiently learn multi-modal distributions that are observed in various applications. A generated model represents the fleet-level CbM model as a combination between two model components: 1) S sets of sparse mixture weights representing individuality of the assets in the fleet; and 2) One set of sparse GGMs that are shared with the S assets to represent commonality across the S assets.

Managing anomaly detection models for fleets of industrial equipment

A method for detecting early indications of equipment failure in an industrial system includes receiving sensor training data collected from industrial equipment under normal conditions and identifying periods of time in the sensor training data when the equipment was functioning normally; finding a pattern for each identified period of time to initialize a plurality of mixture models; learning weighting factors, mean and variance of each of the plurality of mixture models, and removing unimportant models from the plurality of mixture models; determining a Gaussian Markov random field model from surviving mixture models by calculating gating functions for each of the variables and individual mixture models; determining a threshold value of an anomaly score for each variable from the sensor training data; and deploying the model to monitor sensor data from industrial equipment using the threshold values to detect anomalous sensor data values indicative of an impending system failure.

Diagnostic fault detection using multivariate statistical pattern library

A method includes capturing multivariate time-series data comprising two or more data sets from a system captured over a past time period and a present time period, applying at least two sliding time windows to the multivariate time-series data in determining respective data matrices, computing an orthonormal matrix for each of the data matrices, wherein the orthonormal matrix is a signature of fluctuation patterns of a respective data matrix, computing a difference between at least two of the data sets in the past and the present time periods through the orthonormal matrices, and detecting a fault in at least one of the systems by comparing the difference to a threshold.

Change detection using directional statistics

A method providing an analytical technique introducing label information into an anomaly detection model. Effective utilization of label information is based on introducing the degree of similarity between samples. Assuming, for example, there is a degree of similarity between normally labeled samples and no similarity between normally labeled and abnormally labeled samples. Also each sensor value is generated by the linear sum of a latent variable and a coefficient vector specific to each sensor. However, the magnitude of observation noise is formulated to vary according to the label information for the sensor values, and set so that normal label&#x2264;unlabeled&#x2264;anomalously labeled. A graph Laplacian is created based on the degree of similarity between samples, and determines the optimal linear transformation matrix according to a gradient method. A optimal linear transformation matrix is used to calculate an anomaly score for each sensor in the test samples.

Anomaly detection method, program, and system

A method providing an analytical technique introducing label information into an anomaly detection model. The method includes the steps of: inputting measurement data having an anomalous or normal label and measurement data having no label as samples; determining a similarity matrix indicating the relationship between the samples based on the samples; defining a penalty based on the similarity matrix and calculating parameters in accordance with an updating equation having a term reducing the penalty; and calculating a degree of anomaly based on the calculated parameters. The present invention also provides a program and system for detecting an anomaly based on measurement data.

A method, apparatus and computer program for detecting occurrence of an anomaly. The method can exclude arbitrariness and objectively judge whether a variation of a physical quantity to be detected is abnormal or not even when an external environment is fluctuating. The method includes acquiring multiple primary measurement values from a measurement target. Further, calculating and a reference value for each of the multiple primary measurement values by optimal learning. The method further includes calculating a relationship matrix which indicates mutual relationships between the multiple secondary measurement values. Further the method includes calculating an anomaly score for each of the secondary measurement value which indicates the degree of the measurement target being abnormal. The anomaly score is calculated by comparing the secondary measurement value with a predictive value which is calculated based on the relationship matrix and other secondary measurement values.

Detecting occurrence of abnormality

A method providing an analytical technique introducing label information into an anomaly detection model. Effective utilization of label information is based on introducing the degree of similarity between samples. Assuming, for example, there is a degree of similarity between normally labeled samples and no similarity between normally labeled and abnormally labeled samples. Also each sensor value is generated by the linear sum of a latent variable and a coefficient vector specific to each sensor. However, the magnitude of observation noise is formulated to vary according to the label information for the sensor values, and set so that normal label&#x2266;unlabeled&#x2266;anomalously labeled. A graph Laplacian is created based on the degree of similarity between samples, and determines the optimal linear transformation matrix according to a gradient method. A optimal linear transformation matrix is used to calculate an anomaly score for each sensor in the test samples.

An information processing apparatus, a calculation method, a program, and a storage medium for generating a uniformly distributed discrete pattern. To calculate a spatial arrangement of a plurality of elements of a discrete pattern, the plurality of elements being arranged in a spatially discrete manner, an information processing apparatus according to the present invention determines, for each of the elements, a density in an initial position given to the element from a density distribution of the elements in a region where the elements are arranged in the discrete pattern and places, for the initial position of each of the elements, a figure having a size corresponding to the density and representing a region where the element repels other elements and a movement range of the figure. The information processing apparatus minimizes an objective function, computes an optimal solution, and outputs the optimal solutions.

Information processing apparatus, calculation method, program, and storage medium

Provided are a method, an apparatus and a computer program for calculating a risk assessment value for an event sequence, which are capable of calculating the risk assessment value of each even sequence by calculating a totally ordered set on the basis of a partially ordered set indicating the event sequence. The risk assessment value of an event sequence that is a partially ordered set indicating some events of an event group of M kinds of events (M is a finite natural number) in a time series. The partially ordered set is converted into an approximate totally ordered set, and an M-dimensional feature vector is calculated based on the totally ordered set obtained by the conversion. A projection matrix for calculating the risk assessment value is calculated using the calculated M-dimensional feature vector.

Calculating risk assessment value of event sequence

Growing community of inventors

Kawasaki, Japan

Tsuyoshi Ide