Abstract: A method for extracting text information from images includes obtaining an extraction request associated with live data comprising an image; generating, using a prediction model, rotational variant features and rotational invariant features associated with the live data; generating, using the prediction model, text embeddings associated with the rotational variant features using overlapping kernel-based embedding on the live data; generating, using the prediction model, attention values for each pixel in the live data using context attention; applying a trained language model to the text embeddings, attention values, and the live data to generate predictions; and performing extraction actions based on the predictions.

Abstract: A method for controlling an active harmonic filter of an inverter-based resource includes receiving, via a maximum compensation tracker module, a grid feedback signal, determining, via the maximum compensation tracker module, a phase shift signal based, at least in part, on the grid feedback signal, applying, via the maximum compensation tracker module, a phase shift offset signal to the phase shift signal to obtain a modified phase shift signal, determining, via the maximum compensation tracker module, a voltage reference signal for the active harmonic filter based, at least in part, on the grid feedback signal and the modified phase shift signal; and controlling, via the maximum compensation tracker module, the active harmonic filter using the voltage reference signal, wherein the phase shift offset signal ensures that the active harmonic filter injects a current substantially out of phase of a targeted harmonic.

Abstract: Systems and methods for detecting hardware faults in computer-based feedback control systems. Multiple instances of the system control program(s) are run on system processors. System sensor data are input to each instance, and the control commands output by each instance are compared. As instantiations of the same programs receive largely the same sensor data, differences between output commands may indicate the presence of one or more hardware faults.

Abstract: Mechanisms are provided for migrating an application to a new cloud computing system. A causal model is generated based on configuration parameters for a first cloud computing system, monitoring data collected for an execution of the application in the first cloud computing system, and an inserted causal layer. Chaos engineering logic is executed on the causal model to perform a fault injection on the configuration parameters to emulate a second cloud computing system configuration. A mapping, by the causal layer, of the configuration parameters to the monitoring data is learned based on the fault injection. An artificial intelligence for information technology operations (AIOps) model is updated, based on the learned mapping of the causal layer, for monitoring the application execution in the new cloud computing system. The updated AIOps model is provided to an observability tool executing on the new cloud computing system.

Abstract: Mechanisms are provided that detect an anomaly in performance of a hybrid application based on a specification of required performance and collected passive monitoring data, and that generate a causal generative model based on relationships between hybrid application components and computing system architecture components extracted from the passive monitoring data. Root cause identification (RCI) logic is executed on the causal generative model to identify a set of candidate root causes of the detected anomaly. One or more probes are identified for active monitoring data collection targeting the identified set of candidate root causes, which are then executed to collect probe data. Reinforcement learning is performed of the RCI logic to update the RCI logic based on the probe data. The set of candidate root causes is updated based on the reinforcement learning.

Abstract: A method, system, and computer program product for estimating propagation time for an injected fault are configured to: determine normal execution times of respective services in a call graph of an application; determine normal execution times of respective network communications between ones of the services; determine faulty execution times of respective ones of the services; and generate a propagation time for a particular type of fault injected at a particular fault injection location in the call graph based on the determined normal execution times of respective services, the determined normal execution times of respective network communications, and the determined faulty execution times of respective ones of the services.

Abstract: A method for analyzing a PDF document includes: modifying, by an analyzer, a background color of the document to a second color to generate a modified document; converting, by a noise removal module (NRM), the modified document into an image; making, by the NRM, a first determination that the image comprises a background noise, in which the noise is reduced by modifying an area within the image; generating, by a fraud detection module (FDM), at least one image patch from the image; generating, by the FDM, an embedding for the at least one image patch, in which the embedding is processed using a convolutional neural network (CNN) layer; making, by the FDM and based on an output of the CNN layer, a second determination that the document is invalid; and notifying, by the FDM, a user about the second determination using a GUI of the client.

Abstract: A method for automatically responding to a user input includes receiving the user input. The method also includes identifying a current domain associated with the user input. Further, the method includes determining, using a previously trained learning model, a first belief state, where the first belief state is based on the current domain. In addition, the method includes determining, using a reinforcement learning model, a second belief state, where the second belief state is based on the current domain and a reward information. Moreover, the method includes determining an action based on the user input and one selected from the group consisting of the first belief state and the second belief state. Also, the method includes generating a response based on the action and presenting the response to a user.

Abstract: In an approach to improve enhancing the provisioning cloud resources, embodiments receive a set of potential cloud resource providers and predict a performance for a distributed workload on each potential provider of the set according to a machine learning model. Additionally, embodiments inject fault into the set of potential cloud resource providers and measure an impact of the injected fault upon system performance. Embodiments utilize the injected fault to create a system intervention to ensure that the system intervention is carried out on a network in predetermined system. Further, embodiments provision the distributed workload among the potential cloud resource providers according to dynamic conditions output by the set of potential cloud resources based on the measured impact of the injected fault.

Abstract: A method for information extraction from an image-based asset includes: generating, by an encoder, at least one image patch from the asset; generating, by the encoder, an input embedding for the at least one image patch; generating, by the encoder, an output embedding based on the input embedding; inferring, by a decoder, a detail of the image-based asset based on the output embedding and a formatted asset, in which the output embedding is sent by the encoder, wherein the formatted asset is sent by a parser; generating, by the decoder, a decoder output based on the detail, in which the detail comprises at least a feature and a second feature; converting, by a converter, the decoder output into an output asset, in which the decoder output is sent by the decoder; and sending, by the converter, the output asset to a user using a graphical user interface (GUI).

Abstract: Systems and methods for detecting hardware faults in computer-based feedback control systems. Multiple instances of the system control program(s) are run on system processors. System sensor data are input to each instance, and the control commands output by each instance are compared. As instantiations of the same programs receive largely the same sensor data, differences between output commands may indicate the presence of one or more hardware faults.

Abstract: Techniques described herein relate to a method for generating address recommendations. The method includes obtaining, by a recommendation system, an address recommendation request associated with an address, wherein the address is associated with a user; in response to obtaining the address recommendation request: generating a context vector associated with the address; generating an address recommendation based on the context vector using a recommendation model; obtaining user feedback associated with the address recommendation; generating a reward based on the user feedback; and updating the recommendation model based on the context vector, the address recommendation and the reward.

Abstract: Techniques described herein relate to a method for performing knowledge extraction and noise removal for prediction models. The method includes obtaining, by a prediction system, live tabular data; in response to obtaining live tabular data: performing data preprocessing on the live tabular data to generate processed live tabular data; generating a knowledge vector based on the processed live tabular data using a dimensionality reduction model and a tabular attention model; generating a prediction using a prediction model and the knowledge vector; and providing the prediction to a client; wherein the client performs prediction processing using the prediction.

Abstract: Techniques described herein relate to a method for performing capacity planning services. The method includes obtaining a current CP state from a client; in response to obtaining the current state: selecting an action based on the current CP state; providing the action to the client, wherein the client performs the action; in response to providing the action: obtaining a new CP state and a headcount associated with the action; calculating a reward based on the headcount and a reward formula; storing the current CP state, the action, the new CP state, and the reward as a learning set in storage comprising a plurality of learning sets; and performing a learning update using a portion of the plurality of learning sets to generate an updated actor, an updated critic, an updated target actor, and an updated target critic.

Abstract: Methods, apparatus, and processor-readable storage media for localized anomaly detection in digital documents are provided herein. An example computer-implemented method includes identifying at least one image in a digital document; processing the identified at least one image using at least an image compression algorithm; applying a first machine learning model to the at least one processed image, wherein the first machine learning model is trained to detect whether the at least one image comprises one or more modifications; in response to detecting that the at least one image comprises at least one modification, applying a second machine learning model to identify a location in the at least one image corresponding to the at least one modification; and generating an indication that identifies the location of the at least one modification in the at least one image.

Abstract: Methods, apparatus, and processor-readable storage media for automated request processing using an ensemble machine learning framework are provided herein. An example computer-implemented method includes aggregating interaction data associated with a request; computing a weighted score for the request, wherein the weighted score comprises a first component that is based at least in part on a comparison of the aggregated interaction data to a set of keywords and a second component corresponding to a sentiment predicted by a first machine learning model for at least a portion of the aggregated interaction data; using a second machine learning model to determine whether the request is anomalous based at least in part on the weighted score; and in response to determining that the request is anomalous, initiating one or more automated actions for the request.

Abstract: Methods, apparatus, and processor-readable storage media for character-based representation learning for table data extraction using artificial intelligence techniques are provided herein.

Abstract: Techniques described herein relate to a method for generating address recommendations. The method includes obtaining, by a recommendation system, an address recommendation request associated with an address, wherein the address is associated with a user; in response to obtaining the address recommendation request: generating a context vector associated with the address; generating an address recommendation based on the context vector using a recommendation model; obtaining user feedback associated with the address recommendation; generating a reward based on the user feedback; and updating the recommendation model based on the context vector, the address recommendation and the reward.

Abstract: Techniques described herein relate to a method for performing knowledge extraction and noise removal for prediction models. The method includes obtaining, by a prediction system, live tabular data; in response to obtaining live tabular data: performing data preprocessing on the live tabular data to generate processed live tabular data; generating a knowledge vector based on the processed live tabular data using a dimensionality reduction model and a tabular attention model; generating a prediction using a prediction model and the knowledge vector; and providing the prediction to a client; wherein the client performs prediction processing using the prediction.

Abstract: Techniques described herein relate to a method for performing preprocessing of documents for data extraction. The method includes obtaining a document preprocessing request; in response to obtaining a document preprocessing request: obtaining a document associated with the document preprocessing request; performing data preparation on the document to generate an updated document; generating a document type prediction using the updated document and a document type prediction model; identifying a data extraction service of a plurality of data extraction services associated with the document type prediction; and initiating further processing of the document to perform data extraction using the identified data extraction service.