Abstract: Methods and systems for evaluating and enhancing a neural network model include constructing a surrogate model that corresponds to a target neural network model, based on a degree of knowledge about the target neural network model. Adversarial attacks against the surrogate model are generated, based on an attack goal, a level of attacker capability, and an attack model. The target neural network model is tested for accuracy under the generated adversarial attacks to determine a degree of robustness of the target neural network. Robustness of the target neural network model is enhanced by replacing facial occlusions in input images before applying the input images to the target neural network.

Abstract: A method for managing data of dialysis patients by employing a Deep Dynamic Gaussian Mixture (DDGM) model to forecast medical time series data is presented. The method includes filling missing values in an input multivariate time series by model parameters, via a pre-imputation component, by using a temporal intensity function based on Gaussian kernels and multi-dimensional correlation based on correlation parameters to be learned and storing, via a forecasting component, parameters that represent cluster centroids used by the DDGM to cluster time series for capturing correlations between different time series samples.

Abstract: Methods and systems for enhancing a neural network include detecting an occlusion in an input image using a trained occlusion detection neural network. The detected occlusion is replaced in the input image with a neutral occlusion to prevent the detected occlusion from frustrating facial recognition to generate a modified input image. Facial recognition is performed on the modified input image using a trained facial recognition neural network.

Abstract: Methods and systems for training and deploying a neural network mode include training a modular encoder model using training data collected from heterogeneous system types. The modular encoder model includes layers of neural network blocks and a selectively enabled connections between neural network blocks of adjacent layers. Each neural network block includes neural network layers. The modular encoder model is deployed to a system corresponding to one of the heterogeneous system types.

Abstract: A method for acquiring skills through imitation learning by employing a meta imitation learning framework with structured skill discovery (MILD) is presented. The method includes learning behaviors or tasks, by an agent, from demonstrations: by learning to decompose the demonstrations into segments, via a segmentation component, the segments corresponding to skills that are transferrable across different tasks, learning relationships between the skills that are transferrable across the different tasks, employing, via a graph generator, a graph neural network for learning implicit structures of the skills from the demonstrations to define structured skills, and generating policies from the structured skills to allow the agent to acquire the structured skills for application to one or more target tasks.

Abstract: A method for unsupervised multivariate time series trend detection for group behavior analysis is presented. The method includes collecting multi-variate time series data from a plurality of sensors, learning piecewise linear trends jointly for all of the multi-variate time series data, dividing the multi-variate time series data into a plurality of time segments, counting a number of up/down trends in each of the plurality of time segments, for a training phase, employing a cumulative sum (CUSUM), and, for a testing phase, monitoring the CUSUM for trend changes.

Abstract: A computer-implemented method is provided for computer intrusion detection. The method includes establishing a mapping from low-level system calls to user functions in computer programs. The user functions run in a user space of an operating system. The method further includes identifying, using a search algorithm inputting the mapping and a system-call trace captured at runtime, any of the user functions that trigger the low-level system calls in the system-call trace. The method further includes performing, by a processor device, intrusion detection responsive to a provenance graph with program contexts. The provenance graph has nodes formed from the user functions that trigger the low-level system calls in the system-call trace. Edges in the provenance graph have edge labels describing high-level system operations for low-level system call to high-level system operation correlation-based intrusion detection.

Abstract: A method trains a recursive reasoning unit (RRU). The method receives a graph for a set of words and a matrix for a different set of words. The graph maps each word in the set of words to a node with node label and indicates a relation between adjacent nodes by an edge with edge label. The matrix indicates word co-occurrence frequency of the different set of words. The method discovers, by the RRU, reasoning paths from the graph for word pairs by mapping word pairs from the set of words into a source word and a destination word and finding the reasoning paths therebetween. The method predicts word co-occurrence frequency using the reasoning paths. The method updates, responsive to the word co-occurrence frequency, model parameters of the RRU until a difference between a predicted and true word occurrence are less than a threshold amount to provide a trained RRU.

Abstract: Systems and methods for automated software test design and implementation. The system and method being able to establish an initial pool of test cases for testing computer code; apply the initial pool of test cases to the computer code in a testing environment to generate test results; preprocess the test results into a predetermined format; extract metadata from the test results; generate a training sequence; calculate a reward value for the pool of test cases; input the training sequence and reward value into a reinforcement learning agent; utilizing the value output from the reinforcement learning agent to produce a ranking list; prioritizing the initial pool of test cases and one or more new test cases based on the ranking list; and applying the prioritized initial pool of test cases and one or more new test cases to the computer code in a testing environment to generate test results.

Abstract: Methods and systems for training a neural network include collecting model exemplar information from edge devices, each model exemplar having been trained using information local to the respective edge devices. The collected model exemplar information is aggregated together using federated averaging. Global model exemplars are trained using federated constrained clustering. The trained global exemplars are transmitted to respective edge devices.

Abstract: A method performs actions based on business need matching. A set of business need documents are filtered for relevance with respect to a query business need document to remove irrelevant documents based on business need relevance criteria. Hidden business intentions in remaining business need documents are extracted from the set after the filtering. For the query document with respect to the remaining business need documents, the following are computed: a business intention-based matching score, a business entity-based matching score, and an action modeling based matching score. Using an ensemble method, the scores are integrated into a final score, where higher scoring ones of the remaining business need documents more match a business need of the query business need document. Using an automated manufacturing system, a hardware item is co-manufactured responsive to a joint manufacturing venture derived from the final score.

Abstract: Provided is an image processing device, image display device, and program that allow an image captured by an imaging device not having a vibration suppression function or an image whose vibration has been suppressed incompletely to be displayed on a display device with the vibration suppressed. The image processing device includes a motion estimator configured to estimate the amount of motion of an object between a first image and a second image later than the first image and a motion compensator configured to perform a conversion process on the second image so that vibration of the object between the first image and the second image is suppressed, on the basis of the amount of motion of the object. The motion estimator has a first estimation mode in which the amount of motion of the object is estimated in a predetermined search area and a second estimation mode in which the amount of motion of the object is estimated in a larger area than the search area in the first estimation mode.

Abstract: A process for preparing a glass product with marking and the glass product with marking obtained according to the process thereof are described. The process includes 1) coating an ink composition onto a surface of a glass substrate, and 2) heating the glass substrate obtained in step 1). The obtained glass product includes a marking that contains particles having a size of from 150 to 600 nm.

Abstract: A method for training a predictive model includes training a dual-channel neural network model, which includes a static channel to process static information and a dynamic channel to process temporal information, to generate a probability score that characterizes a likelihood of a health event occurring during a dialysis procedure, based on static profile information and temporal measurement information. An augmented model is trained to generate an importance score associated with the probability score, based on the static profile information and the temporal measurement information.

Abstract: Methods and systems for detecting and correcting anomalies include ranking sensors in a cyber-physical system according to a degree of influence each sensor has on a measured performance indicator in the cyber-physical system. An anomaly is detected in the cyber-physical system based on the measured performance indicator. A corrective action is performed responsive to the detected anomaly, prioritized according to sensor rank.

Abstract: Systems and methods for retrieving similar multivariate time series segments are provided. The systems and methods include extracting a long feature vector and a short feature vector from a time series segment, converting the long feature vector into a long binary code, and converting the short feature vector into a short binary code. The systems and methods further include obtaining a subset of long binary codes from a binary dictionary storing dictionary long codes based on the short binary codes, and calculating similarity measure for each pair of the long feature vector with each dictionary long code. The systems and methods further include identifying a predetermined number of dictionary long codes having the similarity measures indicting a closest relationship between the long binary codes and dictionary long codes, and retrieving a predetermined number of time series segments associated with the predetermined number of dictionary long codes.

Abstract: Methods and systems of training and using a neural network model include training a time series embedding model and a text embedding model with unsupervised clustering to translate time series and text, respectively, to a shared latent space. The time series embedding model and the text embedding model are further trained using semi-supervised clustering that samples training data pairs of time series information and associated text for annotation.

Abstract: Systems and methods for predicting system device failure are provided. The method includes performing graph-based predictive maintenance (GBPM) to determine a trained ensemble classification model for detecting maintenance ready components that includes extracted node features and graph features. The method includes constructing, based on testing data and the trained ensemble classification model, an attributed temporal graph and the extracted node features and graph features. The method further includes concatenating the extracted node features and graph features. The method also includes determining, based on the trained ensemble classification model, a list of prediction results of components that are to be scheduled for component maintenance.

Abstract: Systems and methods for determining a remaining useful life of a system. The system and method include one or more processors; a memory coupled to the one or more processors; a data acquisition unit configured to receive run-to-failure time series data; a neural network training unit configured to train a neural network model to determine a point in time that a health index changes from a healthy stage to a degradation stage; a remaining useful life estimation unit configured to estimate a first remaining useful life of the system based on the point in time; estimate a second remaining useful life of the system by converting a feature representation output by the second neural network; minimize the difference between the first remaining useful life and the second remaining useful life; classify the health stage based on a probability; and an output unit configured to send a warning to a user.

Abstract: Systems and methods for early anomaly prediction on multi-variate time series data are provided. The method includes identifying a user labeled abnormal time period that includes at least one anomaly event. The method also includes determining a multi-variate time series segment of multivariate time series data that occurs before the user labeled abnormal time period, and treating, by a processor device, the multi-variate time series segment to include precursor symptoms of the at least one anomaly event. The method includes determining instance sections from the multi-variate time series segment and determining at least one precursor feature vector associated with the at least one anomaly event for at least one of the instance sections based on applying long short-term memory (LSTM). The method further includes dispatching predictive maintenance based on the at least one precursor feature vector.