Abstract: Methods, systems, and computer program products for a joint prediction and improvement framework for machine learning models are provided herein. A method includes obtaining a machine learning model initialized with a set of parameters; identifying one or more actions based on test inputs corresponding to the machine learning model and historical actions related to a task, where the historical actions are dependent on respective historical outputs of the machine learning model; using the identified one or more actions to jointly compute: one or more first values corresponding to inference loss for the machine learning model; and one or more second values based at least in part on a computing cost function associated with the task; and updating the set of parameters of the machine learning model based on the one or more first values and the one or more second values.

Abstract: Embodiments are provided for ticket-agent matching and agent skillset development. In some embodiments, a system includes a processor that executes computer-executable components stored in memory. The computer-executable components can include a matching component that determines, using a ticket profile and a space of agent profiles, a ticket-agent pair including a ticket identifier of a service request and an agent identifier of a particular agent within a pool of agents. The computer-executable components also can include a rematching component that assigns a second agent identifier to the service request to develop a skillset of a second particular agent within the pool of agents, the second agent identifier being associated with an unsatisfactory skill score for a defined skill to resolve the service request.

Abstract: A computer-implemented method, system and computer program product for generating a user interface. A sketch (e.g., wireframe) of a portion of a user interface is received. The sketch is analyzed to predict a set of intended sketches using artificial intelligence based on historical data and/or the user's asset library. A set of intended final sketch renderings of the user interface is then generated and displayed using the set of predicted intended sketches based on historical data or a model trained to extract visual characteristics from existing user interface screens. If the user selects one of the intended final sketch renderings of the user interface as being directed to the intended design of the user interface and indicates that the selected intended final sketch rendering of the user interface corresponds to the final intended design, then code is generated to render the selected final sketch rendering of the user interface.

Abstract: A computer implemented method for selecting service providers includes receiving a set of client requirements and analyzing available service providers based on the received set of client requirements. The method additionally includes scoring the available service providers based on the analysis. The method further includes identifying one or more unstructured external data sources corresponding to the available service providers and analyzing the reliability of the one or more unstructured external data sources with respect to the available service providers. The method further includes adjusting the scoring of the service providers based, at least in part, on the data source reliability, and subsequently providing an optimal selection of service providers based on the adjusted scoring. A computer program product and computer system corresponding to the method are also disclosed.

Abstract: Systems, methods, and computer program products for providing personalized recommendations of devices for monitoring and/or managing a health condition are disclosed, and generally include receiving first structured information regarding a patient and a first set of one or more patient populations; receiving unstructured information regarding at least the patient and a second set of one or more patient populations; analyzing the unstructured information to derive second structured information; determining one or more health metrics to be monitored for the patient based on analyzing each of the first structured information and the second structured information, using a classification model; and determining an optimum set of devices to be used for monitoring the one or more health metrics. In some embodiments, metrics may be continuously monitored to detect a change exceeding an event trigger threshold, and a new set of recommended devices may be generated.

Abstract: In a method for determining anomalous behavior of a candidate taking an exam, a processor receives first exam interface input values captured during an exam session on a candidate testing device. A processor generates a first interaction vector from the first exam interface input values. A processor generates a first interaction timeline from the first interaction vector. A processor determines an anomalous behavior based on a relationship between the first interaction timeline and a selected classification cluster.

Abstract: Computer systems, methods and program products for automating pseudonymization of personal identifying information (PII) using machine learning, metadata, and crowdsourcing patterns to identify and replace PII. Machine learning models are trained for classifying known column names or key names for processing, using metadata. Column or key names are classified to be unprocessed, anonymized or pseudonymized by a pseudonymizer without revealing PII or scrubbing data into a useless format. A library of crowdsourced patterns are utilized for matching PII to data values within column or key names and PII is mapped to replacement methods. Feedback from user annotations retrains the algorithms to improve classification accuracy and Deep Learning algorithms automate the identification of PII using regular expression generation to concisely articulate how pseudonymizers search for PII patterns within a data set.

Abstract: A computer-implemented method according to one embodiment includes receiving input data for a plurality of entities and locations; defining constraints for a group-based allocation model; mapping a distance matrix to a plurality of distance levels to obtain a distance-level formulation; defining a group-level objective function for the group-based allocation model; applying the distance-level formulation to the group-level objective function; solving the group-based allocation model to obtain a group-level assignment; and mapping the group-level assignment to an entity-level assignment.

Abstract: One embodiment provides a method comprising receiving training data and experts' intuition, training a machine learning model based on the training data, predicting a class label for a new data input based on the machine learning model, estimating a degree of similarity of a target attribute of the new data input relative to the training data, and selectively applying a correction to the class label for the new data input based on the degree of similarity prior to providing the class label as an output. The target attribute is an attribute related to the experts' intuition.

Abstract: An approach for providing prediction and optimization of an adversarial machine-learning model is disclosed. The approach can comprise of a training method for a defender that determines the optimal amount of adversarial training that would prevent the task optimization model from taking wrong decisions caused by an adversarial attack from the input into the model within the simultaneous predict and optimization framework. Essentially, the approach would train a robust model via adversarial training. Based on the robust training model, the user can mitigate against potential threats by (adversarial noise in the task-based optimization model) based on the given inputs from the machine learning prediction that was produced by an input.

Abstract: A method of using a computing device to compare performance of multiple algorithms. The method includes receiving, by a computing device, multiple algorithms to assess. The computing device further receives a total amount of resources to allocate to the multiple algorithms. The computing device additionally assigns a fair share of the total amount of resources to each of the multiple algorithms. The computing device still further executes each of the multiple algorithms using the assigned fair share of the total amount of resources. The computing device additionally compares the performance of each of the multiple based on at least one of multiple hardware relative utility metrics describing a hardware relative utility of any given resource allocation for each of the multiple algorithms.

Abstract: Systems and methods for generating user-centric and event-sensitive text summaries are described. For example, summaries may be generated based on user selected reading parameters and user workflow. According to some embodiments, a reinforcement learning module is used to modify a change summarization network based on user feedback. For example, a text summary may change in real-time based on changes to the reader or event context. In some cases, user actions and feedback (e.g., a number of edits to a text summary or the editing time taken by a user) are used to improve prediction of future summaries.

Abstract: Embodiments are provided for ticket-agent matching and agent skillset development. In some embodiments, a system includes a processor that executes computer-executable components stored in memory. The computer-executable components can include a matching component that determines, using a ticket profile and a space of agent profiles, a ticket-agent pair including a ticket identifier of a service request and an agent identifier of a particular agent within a pool of agents. The computer-executable components also can include a rematching component that assigns a second agent identifier to the service request to develop a skillset of a second particular agent within the pool of agents, the second agent identifier being associated with an unsatisfactory skill score for a defined skill to resolve the service request.

Abstract: A method, a computer program product, and a system for predicting low-frequency sensor signal predictions using a hierarchical prediction model. The method includes receiving a historical dataset of high-frequency sensor signal data and low-frequency sensor signal data. The method also includes generating a Gaussian process regression model using the historical dataset and sensor parameters and outputting high-frequency sensor signal predictions. The method also includes generating a hierarchical Gaussian process model using the historical dataset and the high-frequency sensor signal predictions and predicting low-frequency sensor signal predictions.

Abstract: A method of using a computing device executing to interrelate two or more corpuses of dissimilar data that includes receiving input data from each of two or more corpuses of dissimilar data. The computing device computes a pass for each of the input data into two or more encoder-decoder models. The computing device further obtains a prediction of an identity mapping for each of different domains of knowledge from each of the two or more encoder-decoder models. The computing device additionally computes a distribution distance metric as an output from each of a low-dimensional embedding vector representation from each of the two or more encoder-decoder models. The computing device still further computes a function based on each of the predictions from each of the two or more encoder-decoder models and the distribution distance metrics. The computing device additionally updates the two or more encoder-decoder models.

Abstract: One embodiment provides a method comprising receiving data relating to a tenant utilizing a cloud computing environment, and determining one or more classifications for a variation in current workload resource consumption of the tenant based on the data. The current workload resource consumption is indicative of current usage of one or more computing resources of the cloud computing environment.

Abstract: A method for predicting service requests volume includes generating a machine learning model predicting a number of service requests in time series data, based upon a plurality of actually received service requests in the time series data. The method recommends service request features for use in predicting the service requests volume. The method receives a determination from an human-in-the-loop indicating whether the generated machine learning model correctly predicts the number of service requests in time series data, based on the plurality of actually received service requests in the time series data and the recommended service request features. The method selectively updates the machine learning model predicting the number of service requests in times series data, based upon the determination from the human-in-the-loop. The method predicts, using the updated machine learning model, a number of service requests in time series data incoming during a future time period.

Abstract: A method, a computer program product, and a system for predicting low-frequency sensor signal predictions using a hierarchical prediction model. The method includes receiving a historical dataset of high-frequency sensor signal data and low-frequency sensor signal data. The method also includes generating a Gaussian process regression model using the historical dataset and sensor parameters and outputting high-frequency sensor signal predictions. The method also includes generating a hierarchical Gaussian process model using the historical dataset and the high-frequency sensor signal predictions and predicting low-frequency sensor signal predictions.

Abstract: A computer-implemented method, according to one embodiment, includes: receiving an offer request including one or more desired services, and selecting available offerings, each of which include at least one of the desired services. A determination is made whether available benchmarks exist for each of the at least one desired service included in each of the selected available offerings. For each desired service determined as not having available benchmarks, a draft benchmark is computed for each of a plurality of criteria. A confidence weight is also computed for each of the draft benchmarks. The available benchmarks, the draft benchmarks, and the confidence weights are further used to construct an offer which is submitted in response to the received offer request. Moreover, the draft benchmarks and the corresponding confidence weights are re-computed for each of the respective desired services in response to determining that the submitted offer was not accepted.

Abstract: A system for cognitive prioritization for report generation may include a processor and a memory cooperating therewith. The processor may be configured to accept a request for a new report from a user, the request having a user profile importance associated therewith and generate a predicted completion time for the new report based upon a historical completion time prediction model based upon historical data for prior reports. The processor may be configured to generate a predicted importance of the new report based upon a historical importance prediction model based upon the historical data for prior reports and determine a combined predicted importance based upon the user profile importance and the predicted importance. The processor may also be configured to generate a prioritization of the new report among other reports based upon the predicted completion time and the combined predicted importance and generate the new report based upon the prioritization.