Abstract: An embodiment records execution of a workflow comprising a skill, the recording generating workflow data. An embodiment selects, using the workflow data, an ontology tree having above a threshold amount of similarity to the workflow. An embodiment constructs, using the ontology tree and the workflow data, a first skill tree corresponding to the workflow. An embodiment integrates, into an existing skill tree of an application, the first skill tree, the integrating resulting in an integrated skill tree of the application. An embodiment executes, responsive to an intent requesting execution of the skill using new data, using the integrated skill tree and the new data, the skill.

Abstract: An embodiment senses a thermal event in a data center, responsive to the sensed thermal event, identifies an affected server, and decides to migrate a workload from the affected server to a target server. The embodiment senses a performance metric, a temperature and a proximity from the thermal event for each of a plurality of servers, responsive to the sensed performance metric, the temperature and the proximity, computes a weighted score. The embodiment generates a ranking based on the weighted score of each of the plurality of servers. The embodiment selects the target server from the plurality of servers according to the ranking.

Abstract: One or more processors to add identification information of a device to an initialized augmented device retrieval assistance (ADRA) data structure. Received data from ADRA clients monitoring the device within a location includes a network connection status and a time-based position of the device. The ADRA clients dynamically update the ADRA data structure with the location, a current position of the device, and a timeframe of the data from the monitoring by the ADRA clients. A request for ADR assistance is received that includes an identification of a target device, the timeframe, and a location. Location images are generated that include a location layout or floor plan. A time-spatial axis is rendered and sent to an ADRA client and includes location images and target device position within the respective images associated with a specific timeframe on the time-spatial axis.

Abstract: A magneto-resistive random access memory with segmented bottom electrode includes a magnetic tunnel junction pillar above a first portion of a bottom electrode layer, the first portion of the bottom electrode layer includes a metal region. A sidewall spacer is disposed along sidewalls of the magnetic tunnel junction pillar and above a second portion of the bottom electrode layer including a metal-oxide region. The first portion of the bottom electrode layer composed of the metal region and the second portion of the bottom electrode layer composed of the metal-oxide region form the segmented bottom electrode.

Abstract: A computer-implemented method is provided for model training. The method includes training a second end-to-end neural speech recognition model that has a bidirectional encoder to output same symbols from an output probability lattice of the second end-to-end neural speech recognition model as from an output probability lattice of a trained first end-to-end neural speech recognition model having a unidirectional encoder. The method also includes building a third end-to-end neural speech recognition model that has a unidirectional encoder by training the third end-to-end neural speech recognition model as a student by using the trained second end-to-end neural speech recognition model as a teacher in a knowledge distillation method.

Abstract: A computer-implemented method, according to one approach, includes: sending a notification to a host regarding a first logical unit number (LUN) in response to receiving an indication that a first node will be taken offline. The first LUN that the notification is sent to has a preferred communication path that extends through the first node. Communication paths extending through the first node are blocked and communication paths extending through a second node are promoted. Moreover, in response to determining that existing I/O operations previously assigned to the first node have been completed, the first node is taken offline. I/O operations are thereby sent to the first LUN through the second node while the first node is offline.

Abstract: Techniques for forming respective groups of quantum circuit elements (QCEs) on respective crystalline surfaces of a crystalline dielectric (CD) layer are presented. Vias can be formed in the CD layer. Second QCEs can be formed on a second crystalline surface of the CD layer. A seal layer can be applied to the patterned second metallization layer that forms the second QCEs. A handle wafer can be bonded to the seal layer. The chip stack can be turned over to place a substrate at the top, and handle wafer at the bottom, of the chip stack. The substrate and a buried oxide layer can be removed to expose the first crystalline surface of the CD layer. First QCEs can be formed on the first crystalline surface of the CD layer. A portion of the first QCEs can be coupled or interconnected to a portion of the second QCEs.

Abstract: A system and method for pairing two devices for secure communications. A user selects a first device to pair with a second device. The first and second devices have the ability to securely communicate with each other through the use of encrypted communications. An encryption key is written to the first device and then burned into the encryption module on the first device. A third device is selected to pair with the first device. A second encryption key is written to the first device and then burned into the encryption module.

Abstract: Solutions preparing container images and data for container workloads prior to start times of workloads predicted through workload trend analysis. Local storage space on the node is managed based on workload trends, optimizing local storage of image files without requiring frequent reloading and/or deletion of image files, avoiding network intensive I/O operations when pulling images to local storage by workload scheduling systems. Systems perform collection of historical data including image and workload properties; analyze historical data for workload trends, including predicted start times, image files needed, number of nodes and types of nodes. Based on predicted future workload start times, nodes are selected from an ordered list of node requirements and workload properties.

Abstract: Adaptive verifiable training enables the creation of machine learning models robust with respect to multiple robustness criteria. In general, such training exploits inherent inter-class similarities within input data and enforces multiple robustness criteria based on this information. In particular, the approach exploits pairwise class similarity and improves the performance of a robust model by relaxing robustness constraints for similar classes and increasing robustness constraints for dissimilar classes. Between similar classes, looser robustness criteria (i.e., smaller ?) are enforced so as to minimize possible overlap when estimating the robustness region during verification. Between dissimilar classes, stricter robustness regions (i.e., larger ?) are enforced. If pairwise class relationships are not available initially, preferably they are generated by receiving a pre-trained classifier and then applying a clustering algorithm (e.g., agglomerative clustering) to generate them.

Abstract: Embodiments are related to providing external device communication and localization for virtual reality based equipment using radio-frequency identification (RFID). At least two receivers and a transmitter are used to recognize an external device and determine a location of the external device relative to the headset, based on tags coupled to the external device. A three-dimensional (3D) model is downloaded of the external device based on information received by the at least two receivers from the tags. A location of the external device is matched to the 3D model based on the tags. A virtual image is displayed of the external device corresponding to the location of the external device.

Abstract: A method, system, and computer program product for explaining predictions made by black box time series models. The method may include identifying a black box time series model. The method may also include predicting one or more time instances using the black box time series model. The method may also include selecting a predicted time instance from the predicted data. The method may also include receiving training data for the black box time series model. The method may also include generating a set of white box time series models similar to the black box time series model. The method may also include selecting a preferred white box time series model. The method may also include analyzing behavior of the preferred white box time series model. The method may also include generating an explanation illustrating why the black box time series model forecasted the predicted time instance.

Abstract: Aspects of the invention include techniques for providing a learned join cardinality estimation using a join graph representation. A non-limiting example method includes building a join cardinality estimation model. The model can be built by generating a training query having a known join cardinality, generating an adjacency matrix encoding a join graph of the training query, encoding one side of a diagonal axis of the adjacency matrix, and training the join cardinality estimation model using the encoded adjacency matrix and the known join cardinality. The method includes performing an inference using the join cardinality estimation model. The inference includes a predicted join cardinality for a query. The method includes executing a query execution plan for the query using the predicted join cardinality.

Abstract: Mechanisms are provided for performing a tournament selection process of a computer function. A request is received to execute the computer function on an input vector data structure, where a result of the computer function is provided by executing the tournament selection process. The input vector data structure is received, comprising a plurality of values where each value corresponds to a vector slot. An index vector data structure is received that comprises indices of the vector slots of the input vector. Iteration(s) of the tournament selection process are executed to identify a value in the input vector satisfying a criterion of the computer function. An operation is performed on the index vector data structure to generate an indicator vector data structure that uniquely identifies a slot in the input vector data structure that is a result of the computer function being executed on the input vector data structure.

Abstract: A cache hit-miss prediction is determined for a memory access instruction using a predictor. The predictor includes a tracker for the memory access instruction. The tracker is used to provide a prediction confidence level of the cache hit-miss prediction for the memory access instruction. Using the tracker, the prediction confidence level of the cache hit-miss prediction is ascertained. Based on the prediction confidence level indicating the cache hit-miss prediction is to be used, the cache hit-miss prediction is provided to be used in instruction execution scheduling.

Abstract: A computer-implemented method, a computer program product, and a computer system for parallel cross validation in collaborative machine learning. A server groups local models into groups. In each group, each local device uses its local data to validate accuracies of the local models and sends a validation result to a group leader or the server. The group leader or the server selects groups whose variances of the accuracies are not below a predetermined variance threshold. In each selected group, the group leader or the server compares an accuracy of each local model with an average value of the accuracies and randomly selects one or more local models whose accuracies do not exceed a predetermined accuracy threshold. The server obtains weight parameters of selected local models and updates the global model based on the weight parameters.

Abstract: Systems and techniques that facilitate efficient synthesis of optimal multi-qubit Clifford circuits are provided. In various embodiments, a system can receive as input a number n representing a quantity of qubits. In various instances, the system can generate, via a cost-invariant reduction function, as output a library of different n-qubit canonical representatives that respectively correspond to different cost-invariant equivalence classes of n-qubit Clifford group elements. In various embodiments, a system can receive as input a first Clifford group element. In various aspects, the system can search a database of canonical representatives, wherein different canonical representatives in the database respectively correspond to different cost-invariant equivalence classes of Clifford group elements.

Abstract: The present invention provides techniques for learning Mahalanobis distance similarity metrics from data for individually fair machine learning models. In one aspect, a method for learning a fair Mahalanobis distance similarity metric includes: obtaining data with similarity annotations; selecting, based on the data obtained, a model for learning a Mahalanobis covariance matrix ?; and learning the Mahalanobis covariance matrix ? from the data using the model selected, wherein the Mahalanobis covariance matrix ? fully defines the fair Mahalanobis distance similarity metric.

Abstract: A computer-implemented method, according to one approach, includes: receiving observability data produced by a first portion of a distributed application. The observability data is evaluated and in response to determining that the observability data indicates the first portion of the distributed application is degraded, the amount of the observability data produced by the first portion of the distributed application is increased. Additionally, the amount of observability data produced by other portions of the distributed application that are correlated with the first portion of the distributed application is also increased.

Abstract: A computer implemented method creates microservices for an application. A number of processor units clusters programs and data structures for the application using runtime metadata to form groups of the programs and data structures. The runtime metadata is obtained from running the application. The number of processor units creates a design for the microservices for the application using the groups of the programs and the data structures.