Abstract: Systems and methods described herein which can involve for a first input of a plurality of labeled images of a new domain task, processing the first plurality of labeled images through a plurality of backbone snapshots, each of the backbone snapshots representative of a model trained across a plurality of other domain tasks, each of the plurality of backbone snapshots configured to output a first plurality of features responsive to the input; processing a second input of second plurality of unlabeled images through the plurality of backbone snapshots to output a second plurality of features responsive to the second input; and generating a representative model for the new domain task from the clustering and transformation of the first plurality of features and as associated from the clustered and transformed second plurality of features.

Abstract: Example implementations involve systems and methods to create robust visual inspection datasets and models. The novel method learns and transfers damage representation from few samples to new images. The proposed method introduces a generative region-of-interest based adversarial network with the aim of learning a common damage representation and transferring it to an unseen image. The proposed approach shows the benefit of adding damage-region-based component, since existing methods fail to transfer the damages. The proposed method successfully generated images with variations in context and conditions to improve model generalization for small datasets.

Abstract: Example implementations described herein involve an approach to address an imperfect simulator challenge using off-line data plus reward modification. The proposed solution is robust to simulator error, and therefore, it requires less maintenance in keeping the simulators updated. Even when the simulators are accurate, it is costly to keep them accurate over time. Moreover, compared to other robust reinforcement learning algorithms, the proposed approach does not assume the distribution of uncertainties in the simulator are known. Less complexity leads to fewer potential errors as well as lower computational cost during the training. Finally, the proposed approach has better performance compared to the state-of-the-art methods (higher overall cumulative rewards).

Abstract: Example implementations described herein involve systems and methods for generating an ensemble of deep learning or neural network models, which can involve, for a training set of data, generating a plurality of model samples for the training set of data, the plurality of model samples generated from deep learning or neural network methods; and aggregating output of the model samples to generate an output of the ensemble models.

Abstract: Example implementations involve systems and methods to advance data acquisition systems for automated visual inspection using a mobile camera infrastructure. The example implementations address the uncertainty of localization and navigation under semi-controlled environments. The approach combines object detection models and navigation planning to control the quality of visual inputs in the inspection process. The solution guides the operator (human or robot) to collect only valid viewpoints to achieve higher accuracy. Finally, the learning models and navigation planning are generalized to multiple type and size of inspection objects.

Abstract: Example implementations described herein involve defect analysis for images received from a camera system, which can involve applying a first model configured to determine regions of interest of the object from the images, applying a second model configured to identify localized areas of the object based on the regions of interest on the images; and applying a third model configured to identify defects in the localized ones of the images.

Abstract: Example implementations described herein involve systems and methods for generating an ensemble of deep learning or neural network models, which can involve, for a training set of data, generating a plurality of model samples for the training set of data, the plurality of model samples generated from deep learning or neural network methods; and aggregating output of the model samples to generate an output of the ensemble models.

Abstract: In some examples, a system may receive from a device, speech sound patterns corresponding to a voice input related to equipment. Further, the system may determine an identity of a person associated with the device, and may identify the equipment related to the voice input. Using at least one of the received speech sound patterns or a text conversion of the speech sound patterns, along with an equipment history of the identified equipment, as input to one or more machine learning models, the system may determine, at least partially, an instruction related to the equipment. Additionally, the system may send, to the device, the instruction related to the equipment as an audio file for playback on the device.

Abstract: Example implementations involve a framework for knowledge base construction of components and problems in short texts. The framework extracts domain-specific components and problems from textual corpora such as service manuals, repair records, and public Q/A forums using: 1) domain-specific syntactic rules leveraging part of speech tagging (POS), and 2) a neural attention-based seq2seq model which tags raw sentences end-to-end identifying components and their associated problems. Once acquired, this knowledge can be leveraged to accelerate the development and deployment of intelligent conversational assistants for various industrial AI scenarios (e.g., repair recommendation, operations, and so on) through better understanding of user utterances. The example implementations give better tagging accuracy on various datasets outperforming well known off-the-shelf systems.

Abstract: Example implementations relate to updating an inference graph absent node locking. For example, a processor executing a first thread may receive a first task for updating a node of an inference graph stored by a storage device accessible to a second thread, the first task being assigned during a first iteration of a graph update loop. Absent locking the node from access by the second thread, the processor may generate a value for the node and update the node with the value. Based on detecting that each node of the inference graph has been updated, the processor may continue with a second iteration of the graph update loop.

Abstract: Example implementations involve a framework for knowledge base construction of components and problems in short texts. The framework extracts domain-specific components and problems from textual corpora such as service manuals, repair records, and public Q/A forums using: 1) domain-specific syntactic rules leveraging part of speech tagging (POS), and 2) a neural attention-based seq2seq model which tags raw sentences end-to-end identifying components and their associated problems. Once acquired, this knowledge can be leveraged to accelerate the development and deployment of intelligent conversational assistants for various industrial AI scenarios (e.g., repair recommendation, operations, and so on) through better understanding of user utterances. The example implementations give better tagging accuracy on various datasets outperforming well known off-the-shelf systems.

Abstract: Example implementations described herein involve defect analysis for images received from a camera system, which can involve applying a first model configured to determine regions of interest of the object from the images, applying a second model configured to identify localized areas of the object based on the regions of interest on the images; and applying a third model configured to identify defects in the localized ones of the images.

Abstract: In some examples, a system may receive from a device, speech sound patterns corresponding to a voice input related to equipment. Further, the system may determine an identity of a person associated with the device, and may identify the equipment related to the voice input. Using at least one of the received speech sound patterns or a text conversion of the speech sound patterns, along with an equipment history of the identified equipment, as input to one or more machine learning models, the system may determine, at least partially, an instruction related to the equipment. Additionally, the system may send, to the device, the instruction related to the equipment as an audio file for playback on the device.

Abstract: A technique includes performing graph inference in a graph inference engine that includes multiple processing nodes to determine assignments for vertices of a graph. Performing the graph inference includes controlling remote memory accesses within the engine, including storing first data in a local memory of the first processing node, where the first data represents at least assignments for a plurality of vertices of the graph; in the first processing node, determining updates for the assignments for a subset of the plurality of vertices of a partition of the graph assigned to the first processing node and modifying the first data based on the updates; and communicating the updates to at least one other processing node of the multiple processing nodes, where at least one other partition of the graph is assigned to the other processing node(s).

Abstract: A report indicating a user-reported probability of a successful outcome is received. A behavior and information model is estimated based on the report. The behavior and information model includes a behavior model component having a bias parameter and a consistency parameter. The behavior and information model includes an information model component having a first user-believed probability of a successful outcome and a second user-believed probability of a successful outcome. The behavior and information model is used to yield a model-determined probability of a successful outcome that more accurately reflects a probability of a successful outcome than the user-reported probability of a successful outcome does.

Abstract: Methods, systems, and computer-readable and executable instructions are provided for resource planning. Resource planning can include identifying a resource capacity for a given resource planning horizon, identifying a resource pool guideline that defines a priority of resources from which to satisfy a full-time equivalent (FTE) resource planning requirement, and performing resource planning by satisfying the FTE requirement based on the resource capacity and the resource pool guideline.

Abstract: Partitioning job requirements for resource planning can include identifying resource capacity and matching scores of a resource pool for a resource planning horizon, partitioning a full-time equivalent (FTE) job requirement into a plurality of positions based on the resource capacity and matching scores of the resource pool for the resource planning horizon using a computing device, and performing resource planning by allocating the resource capacity to the plurality of positions

Abstract: Information may be extracted from a document. A new pattern may be identified in the document. Classification may be performed based on the extracted information.

Abstract: There is provided a computer-implemented method of generating a data forecasts for different levels of an entity. The method includes generating an aggregate forecast for an upper level entity comprised of two or more components. The method also includes determining mean values and a coefficient of variation for a probability distribution corresponding to future expected decomposition rates for each of the two or more components. A probability distribution parameter vector is computed based on the mean values and the coefficient of variation. The expected future decomposition rates for each of the two or more components may be computed based on the probability distribution parameter vector and a sample observation corresponding to previously observed decomposition values of each of the two or more components. Component forecasts corresponding to each of the two or more components may be computed based on the aggregate forecast and the expected future decomposition rates.

Abstract: Techniques for generating optimized surgery schedules are described in various implementations. In one example implementation, a method that implements the techniques includes receiving a plurality of surgery scheduling requests for surgical procedures to be performed at a surgical facility having a plurality of operating rooms. The method also includes identifying resource constraints associated with the surgery scheduling requests, and identifying an optimization goal for the surgical facility, the optimization goal being defined using weighted optimization parameters. The method also includes generating a proposed surgery schedule for the surgical facility that includes sequencing and operating room assignments for each of the surgical procedures to be performed, the proposed surgery schedule satisfying the resource constraints and being optimized based on the optimization goal for the surgical facility.