Abstract: Techniques that facilitate matrix factorization associated with graphics processing units are provided. In one example, a system includes a first graphics processing unit, a second graphics processing unit and a central processing unit. The first graphics processing unit processes a first data block of a data matrix associated with a matrix factorization system to generate first information for the matrix factorization system. The second graphics processing unit processes a first portion of a second data block of the data matrix separate from a second portion of the second data block to generate second information for the matrix factorization system. The central processing unit processes a machine learning model for the matrix factorization system based on at least the first information provided by the first graphics processing unit and the second information provided by the second graphics processing unit.

Abstract: A system, method, and computer program product for verifying signatures. The system includes at least one processing component, at least one memory component, and a reference storage comprising a set of reference signatures. The system also includes a model generator configured to generate a signature model based on the set of reference signatures. Further, the system includes a verification component configured to receive a signature, and determine whether the signature is valid.

Abstract: A computer-implemented method includes obtaining data associated with execution of a model deployed in a computing environment. At least a portion of the obtained data are analyzed to detect one or more failure conditions associated with the model. One or more restoration operations are executed to generate one or more restoration results to address one or more detected failure conditions. At least a portion of the one or more restoration results is sent to the computing environment in which the model is deployed.

Abstract: The generating of actionable recommendations for tuning model metrics of an Artificial Intelligence (AI) system includes partitioning a key performance indicator (KPI) range associated with a target system into a plurality of buckets. Log data including at least one KPI of the target system and one or more AI model metrics is partitioned and distributed across the plurality of buckets. For each bucket, an aggregate value of the one or more AI model metrics across the log data is computed and weighted according to the volume of log data in that bucket. A correlation factor between the aggregate value and a representative KPI value for each bucket is determined. A model tuning recommendation to increase ranking of the AI model metrics according to the determined correlation factor is provided to an output device and/or to the AI system for updating the one or more AI model.

Abstract: Techniques that facilitate matrix factorization associated with graphics processing units are provided. In one example, a system includes a first graphics processing unit, a second graphics processing unit and a central processing unit. The first graphics processing unit processes a first data block of a data matrix associated with a matrix factorization system to generate first information for the matrix factorization system. The second graphics processing unit processes a first portion of a second data block of the data matrix separate from a second portion of the second data block to generate second information for the matrix factorization system. The central processing unit processes a machine learning model for the matrix factorization system based on at least the first information provided by the first graphics processing unit and the second information provided by the second graphics processing unit.

Abstract: A computing method receives a labeled sample from an annotator. The method may determine a plurality of reference model risk scores for the first labeled sample, where each reference model risk score corresponds to an amount of risk associated with adding the first labeled sample to a respective reference model of a plurality of reference models. The method may determine an overall risk score for the first labeled sample based on the plurality of reference model risk scores. The method may further determine a probe for confirmation of the first labeled sample and a trust score for the annotator by sending the probe to one or more annotators. In response to determining a trust score for the annotator the method may add the labeled sample to a ground truth or reject the labeled sample.

Abstract: Methods and systems are provided for configurable and non-invasive protection of private information in a user input to a software application that handles real-time information. A method includes detecting, by a filter in real-time, private information in the user input. The method further includes forming, by the filter, a filtered user input from the user input, by maintaining non-private information from the user input in the filtered user input, extracting and encrypting the private information in the user input and attaching the encrypted private information to the filtered user input, and replacing the private information in the user input with unique identifiers in the filtered user input. The unique identifiers are configured to be exploitable by the software application to achieve an intended function of the software application for the user. The method also includes transmitting, by a communications redirector, the filtered user input over a communication channel.

Abstract: In an embodiment, a method for inspecting and transforming a machine learning model includes receiving a request that includes the machine learning model and a configuration object that provides an indication of a selected strategy. In the embodiment, the method includes creating a partially specified task graph that includes a first placeholder node for a future expanded task node. In the embodiment, the method includes performing a dynamic expansion and execution phase that includes, repeatedly (a) using a cognitive engine to evaluate whether to revise the partially specified task graph based at least in part on the selected strategy, and (b) using a processor-based execution engine to perform an action specified by the complete node. In an embodiment, the dynamic expansion and execution phase repeats until after the cognitive engine adds a consolidated results node.

Abstract: A system, apparatus and a method for constructing pipelines, including enumerate a subspace of valid integrated pipelines, collecting a set of metrics for each of the plurality of pipelines, reducing the set of metrics to a single metric, selecting a final integrated pipeline from among the valid integrate pipelines based on reduced metric.

Abstract: Aspects of the present invention disclose a method, computer program product, and system for detecting and mitigating adversarial virtual interactions. The method includes one or more processors initiating a mitigation protocol on interactions between the user and the virtual agent, wherein the mitigation protocol is based on the actions performed by the user while interacting with the virtual agent. The method further includes one or more processors, in response to initiating the mitigation protocol on interactions between the user and the virtual agent, generating a lower fidelity response from the virtual agent to the user, wherein the lower fidelity response is a progressive dilution of the precision of language of an original response from the virtual agent to the user prior to the user exceeding the risk level threshold.

Abstract: Embodiments include method, systems and computer program products for a path-sensitive contextual help system. In some embodiments, user actions are obtained from a user session of a user. A concrete user action trace is captured using the obtained user actions, wherein the concrete user action trace is a subset of user actions from the user session. An abstract user action trace is generated using the concrete user action trace. A help action corresponding to the abstract user action trace is identified and, in some embodiments, is presented to the user.

Abstract: Embodiments include method, systems and computer program products for a path-sensitive contextual help system. In some embodiments, user actions are obtained from a user session of a user. A concrete user action trace is captured using the obtained user actions, wherein the concrete user action trace is a subset of user actions from the user session. An abstract user action trace is generated using the concrete user action trace. A help action corresponding to the abstract user action trace is identified and, in some embodiments, is presented to the user.

Abstract: Aspects of the present invention disclose a method, computer program product, and system for detecting and mitigating adversarial virtual interactions. The method includes one or more processors detecting a user communication that is interacting with a virtual agent. The method further includes one or more processors determining a risk level associated with the detected user communication based on one or more actions performed by the detected user while interacting with the virtual agent. The method further includes one or more processors in response to determining that the determined risk level associated with the detected user communication exceeds a risk level threshold, initiating, a mitigation protocol on interactions between the detected user and the virtual agent, where the mitigation protocol is based on the actions performed by the detected user while interacting with the virtual agent.

Abstract: Computer-implemented methods, program products, and systems for provenance-based defense against poison attacks are disclosed. In one approach, a method includes: receiving observations and corresponding provenance data from data sources; determining whether the observations are poisoned based on the corresponding provenance data; and removing the poisoned observation(s) from a final training dataset used to train a final prediction model. Another implementation involves provenance-based defense against poison attacks in a fully untrusted data environment. Untrusted data points are grouped according to provenance signature, and the groups are used to train learning algorithms and generate complete and filtered prediction models. The results of applying the prediction models to an evaluation dataset are compared, and poisoned data points identified where the performance of the filtered prediction model exceeds the performance of the complete prediction model.

Abstract: Techniques that facilitate matrix factorization associated with graphics processing units are provided. In one example, a system includes a first graphics processing unit, a second graphics processing unit and a central processing unit. The first graphics processing unit processes a first data block of a data matrix associated with a matrix factorization system to generate first information for the matrix factorization system. The second graphics processing unit processes a first portion of a second data block of the data matrix separate from a second portion of the second data block to generate second information for the matrix factorization system. The central processing unit processes a machine learning model for the matrix factorization system based on at least the first information provided by the first graphics processing unit and the second information provided by the second graphics processing unit.

Abstract: A method, system, and/or computer program product manages the lifecycle of trained models used to deliver cognitive services. One or more processors obtain and deploy a cognitive engine that utilizes artificial intelligence (AI), machine learning, and/or similar algorithms. One or more processors obtain and deploy a version of a trained model that includes data that supports cognitive operations of the cognitive engine within a cognitive service. In response to changes to the input used to produce the trained model, one or more processors obtain and deploy a subsequent version of the trained model in support of the cognitive service.

Abstract: Methods and systems are provided for configurable and non-invasive protection of private information in a user input to a software application that handles real-time information. A method includes detecting, by a filter in real-time, private information in the user input. The method further includes forming, by the filter, a filtered user input from the user input, by maintaining non-private information from the user input in the filtered user input, extracting and encrypting the private information in the user input and attaching the encrypted private information to the filtered user input, and replacing the private information in the user input with unique identifiers in the filtered user input. The unique identifiers are configured to be exploitable by the software application to achieve an intended function of the software application for the user. The method also includes transmitting, by a communications redirector, the filtered user input over a communication channel.

Abstract: Techniques that facilitate matrix factorization associated with graphics processing units are provided. In one example, a system includes a first graphics processing unit, a second graphics processing unit and a central processing unit. The first graphics processing unit processes a first data block of a data matrix associated with a matrix factorization system to generate first information for the matrix factorization system. The second graphics processing unit processes a first portion of a second data block of the data matrix separate from a second portion of the second data block to generate second information for the matrix factorization system. The central processing unit processes a machine learning model for the matrix factorization system based on at least the first information provided by the first graphics processing unit and the second information provided by the second graphics processing unit.

Abstract: Performance of a computer implementing a machine learning system is improved by providing, via a graphical user interface, to an annotator, unlabeled corpus data to be labeled; obtaining, via the interface, labels for the unlabeled corpus data; and detecting, with a consistency calculation routine, concurrent with the labeling, internal inconsistency and/or external inconsistency in the labeling. Responsive to the detection, intervene in the labeling with a reactive intervention subsystem until the inconsistency is addressed. The labeling is completed subsequent to the intervention; the system is trained to provide a trained machine learning system, based on results of the completed labeling; and classification of new data is carried out with the trained system. Proactive intervention schemes are also provided.

Abstract: Aspects of the present invention disclose a method, computer program product, and system for detecting and mitigating adversarial virtual interactions. The method includes one or more processors detecting a user communication that is interacting with a virtual agent. The method further includes one or more processors determining a risk level associated with the detected user communication based on one or more actions performed by the detected user while interacting with the virtual agent. The method further includes one or more processors in response to determining that the determined risk level associated with the detected user communication exceeds a risk level threshold, initiating, a mitigation protocol on interactions between the detected user and the virtual agent, where the mitigation protocol is based on the actions performed by the detected user while interacting with the virtual agent.