Abstract: A vision transformer is provided featuring improved computational efficiency. A plurality of image vectors corresponding to an input image is provided to a sequence of neural network layers configured to generate an image classification. The neural network layers comprise at least one scatter layer coupled to at least one attention layer, the scatter layer configured to receive the plurality of image patch vectors and to generate low-frequency tokens and high-frequency tokens by applying a dual-tree complex wavelet transform and applying tensor and Einstein mixing to each set of tokens, respectively. The tokens are thereafter transformed back to the physical domain by an inverse scatter network, provided to a multi-layer perceptron (MLP) layer which in turn provides an output to the at least one attention layer that includes multi-head self-attention and MLP layers further coupled to a classifier head configured to generate a classification of the input image.

Abstract: Methods, systems, apparatuses, and computer-readable storage mediums are descried for identifying a similarity between queries. An intermediate representation generator receives a set of queries from a repository, each query in the set of queries having generated a corresponding set of data stored in a data store. An intermediate representation is generated for each query, where the intermediate representation is characterized by a feature associated with text specified in the query. A similarity determiner determines similarity scores between pairs of intermediate representations. A pair of intermediate representations with a similarity score above a threshold is identified. An indication is generated that sets of data corresponding to queries corresponding to the intermediate representations are overlapping.

Abstract: Methods, systems, apparatuses, and computer-readable storage mediums are descried for identifying a similarity between queries. An intermediate representation generator receives a set of queries from a repository, each query in the set of queries having generated a corresponding set of data stored in a data store. An intermediate representation is generated for each query, where the intermediate representation is characterized by a feature associated with text specified in the query. A similarity determiner determines similarity scores between pairs of intermediate representations. A pair of intermediate representations with a similarity score above a threshold is identified. An indication is generated that sets of data corresponding to queries corresponding to the intermediate representations are overlapping.

Abstract: Methods, systems, apparatuses, and computer-readable storage mediums are descried for identifying a similarity between queries. An intermediate representation generator receives a set of queries from a repository, each query in the set of queries having generated a corresponding set of data stored in a data store. An intermediate representation is generated for each query, where the intermediate representation is characterized by a feature associated with text specified in the query. A similarity determiner determines similarity scores between pairs of intermediate representations. A pair of intermediate representations with a similarity score above a threshold is identified. An indication is generated that sets of data corresponding to queries corresponding to the intermediate representations are overlapping.

Abstract: This application relates to systems and methods for automatically correcting labels in untrusted data based on a small sample of trusted data in a training database. In some examples, training data may be divided into a trusted dataset and an untrusted dataset using stratified sampling. An adversarial algorithm may be used to reassign labels in the data samples associated with the untrusted data based on a set of features in the data and labels in the trusted dataset. The untrusted dataset with the reassigned labels may then be used to train a machine learning model.

Abstract: Described are examples for generating a model for forecasting time series data. For a timeseries data set, one or more layers can be provided, where each layer in the one or more layers includes, for each timeseries data input in at least a portion of multiple timeseries data inputs, generating, for the timeseries data input, a short range output from a causal convolution process that is based on timeseries data inputs from the timeseries data set that are associated with timestamps within a threshold time before the timestamp of the timeseries data input, and generating, for the timeseries data input, a long range output from a transformer process based on the short range outputs from the causal convolution process for each timeseries data input from at least the portion of the multiple timeseries data inputs that are associated with timestamps before the timestamp of the timeseries data input.

Abstract: Example aspects include techniques for anomaly detection via sparse judgmental samples. These techniques may include generating, via lexical analysis, a plurality of tokens from a textual representation of a machine learning (ML) model and generating, via a parser, based on the plurality of tokens, an abstract syntax tree (AST) corresponding to the ML model. In addition, the techniques may include identifying a data dependency of the ML model based on an AST node within the AST, the AST node corresponding to a data source and the data dependency indicating the ML model depends on the data source. Further, the techniques may include detecting a potential issue associated with the data source, and transmitting, based on the data dependency, an alert notification in response to the potential issue.

Abstract: In various examples, a system can obtain a first time series data set, the first time series data set including a plurality of data elements. Each data element can include value data and corresponding time data. Based on the first time series data set, the system can generate a second data set and a third dataset. The second dataset can indicate one or more data elements with missing value data and the third dataset can include extremeness data. The extremeness data can indicate an extremeness score for each data element of the plurality of data elements. Additionally, based on the first time series data set, the second data set and a third dataset, the system can implement a set of operations that generate a substitute value data for each data element of the one or more data elements that is missing value data.

Abstract: This application relates to apparatus and methods for automatically predicting values for a future time period based on time series data of a previous time period. In some examples, a computing device employs multiple algorithms or predictions models to determine baseline predictions and bias predictions accounting for both normal and surge-induced events in the future time period. Accuracy of the algorithms and exogenous variables, such as holidays, events, temporal indicators, are leveraged to accurately predict future values. Baseline predictions using baseline algorithms are aggregated with bias predictions associated with surge events to determine final predictions for the future time period without compromising on the accuracy and efficiency of the predictions for both normal and surge-induced events.

Abstract: This application relates to systems and methods for automatically correcting labels in untrusted data based on a small sample of trusted data in a training database. In some examples, training data may be divided into a trusted dataset and an untrusted dataset using stratified sampling. An adversarial algorithm may be used to reassign labels in the data samples associated with the untrusted data based on a set of features in the data and labels in the trusted dataset. The untrusted dataset with the reassigned labels may then be used to train a machine learning model.

Abstract: This disclosure describes, generally, methods and systems for implementing memory overcommit of virtual machines. The method includes establishing a plurality of virtual machines on a physical machine, broadcasting, from each of the plurality of virtual machines to a central scheduler, resource usage requirements, and then based at least in part on the resource usage requirements broadcasted from each of the plurality of virtual machines, determining a resource requirements schedule for each of the plurality of virtual machines. The method further includes receiving at least one resource request from at least one of the plurality of virtual machines, based on the resource requirements schedule, un-assigning resources from at least one of the plurality of virtual machines, and assigning the un-assigned resources to the at least one of the plurality of virtual machines which initiated the resource request.

Abstract: This disclosure describes, generally, methods and systems for implementing memory overcommit of virtual machines. The method includes establishing a plurality of virtual machines on a physical machine, broadcasting, from each of the plurality of virtual machines to a central scheduler, resource usage requirements, and then based at least in part on the resource usage requirements broadcasted from each of the plurality of virtual machines, determining a resource requirements schedule for each of the plurality of virtual machines. The method further includes receiving at least one resource request from at least one of the plurality of virtual machines, based on the resource requirements schedule, un-assigning resources from at least one of the plurality of virtual machines, and assigning the un-assigned resources to the at least one of the plurality of virtual machines which initiated the resource request.

Abstract: This disclosure describes, generally, methods and systems for implementing memory overcommit of virtual machines. The method includes establishing a plurality of virtual machines on a physical machine, broadcasting, from each of the plurality of virtual machines to a central scheduler, resource usage requirements, and then based at least in part on the resource usage requirements broadcasted from each of the plurality of virtual machines, determining a resource requirements schedule for each of the plurality of virtual machines. The method further includes receiving at least one resource request from at least one of the plurality of virtual machines, based on the resource requirements schedule, un-assigning resources from at least one of the plurality of virtual machines, and assigning the un-assigned resources to the at least one of the plurality of virtual machines which initiated the resource request.

Abstract: According to an aspect of the present invention, nodes for hosting of new virtual machines (VM) are selected according to approaches designed to reduce power consumption in a grid. In an embodiment, the approaches are designed to facilitate the possibility of freeing one or more nodes from hosting VMs to power down the nodes, thereby reducing power consumption. Thus, an example approach is based on provisioning a new VM on a node which currently (immediately prior to provisioning) has the maximum resource consumption. Another example approach is based on provisioning a new VM on a node which currently has small-sized VMs in terms of resource requirements. In yet another embodiment, the approach is based on provisioning a new VM on a node located in a geographical area having low power tariffs.

Abstract: An aspect of the present invention facilitates selecting suitable nodes to host virtual machines (VMs) in an environment containing a large number of nodes (such as a grid). In one embodiment, information indicating corresponding resources available in each machine node (a node capable of hosting VMs) in the grid is maintained distributed over a set of management nodes contained in the grid. On receiving an indication that a VM requiring a set of resources is sought to be hosted, a machine node having available the set of resources is identified based on the distributed information. The VM is then provisioned/hosted on the identified machine node. The maintenance of the resource availability information distributed across multiple management nodes enables the solution to be scaled for use in environments having a large number of nodes.

Abstract: An aspect of the present invention mitigates reduction in availability level during maintenance of nodes in a cluster. In one embodiment, on receiving an indication that a maintenance activity is to be performed on the cluster, a scaling out of the cluster is first performed to add some nodes having the maintenance activity already performed, followed by a scaling in of the cluster to remove some of the nodes in the cluster which do not yet have the maintenance activity performed. The scaling out is performed before any scaling in of the cluster such that the number of nodes available in the cluster after the scaling in is not less than the number of nodes in the cluster at the time of receiving the indication. Accordingly, the reduction in availability level (which is based on the number of nodes available) is mitigated.

Abstract: An aspect of the present invention reduces the recovery time for business organizations in case of disasters. In one embodiment, a disaster recovery system containing a primary site and a backup site (implemented as a cluster) is maintained. Application instances are executed in both the primary site and the backup site, with the number of instances executed on the backup site being fewer than that executed on the primary site. During normal operation, user requests received are processed using only the instances executing in the primary site, while the instances executing in the backup site are used in a standby state. On identifying that a disaster has occurred, the user requests received immediately after identification of the disaster are processed using only the instances executing in the backup site. The cluster at the backup site is then scaled out to add application instances until a desired level/percentage is achieved.

Abstract: This disclosure describes, generally, methods and systems for implementing memory overcommit of virtual machines. The method includes establishing a plurality of virtual machines on a physical machine, broadcasting, from each of the plurality of virtual machines to a central scheduler, resource usage requirements, and then based at least in part on the resource usage requirements broadcasted from each of the plurality of virtual machines, determining a resource requirements schedule for each of the plurality of virtual machines. The method further includes receiving at least one resource request from at least one of the plurality of virtual machines, based on the resource requirements schedule, un-assigning resources from at least one of the plurality of virtual machines, and assigning the un-assigned resources to the at least one of the plurality of virtual machines which initiated the resource request.

Abstract: An aspect of the present invention facilitates selecting suitable nodes to host virtual machines (VMs) in an environment containing a large number of nodes (such as a grid). In one embodiment, information indicating corresponding resources available in each machine node (a node capable of hosting VMs) in the grid is maintained distributed over a set of management nodes contained in the grid. On receiving an indication that a VM requiring a set of resources is sought to be hosted, a machine node having available the set of resources is identified based on the distributed information. The VM is then provisioned/hosted on the identified machine node. The maintenance of the resource availability information distributed across multiple management nodes enables the solution to be scaled for use in environments having a large number of nodes.

Abstract: According to an aspect of the present invention, nodes for hosting of new virtual machines (VM) are selected according to approaches designed to reduce power consumption in a grid. In an embodiment, the approaches are designed to facilitate the possibility of freeing one or more nodes from hosting VMs to power down the nodes, thereby reducing power consumption. Thus, an example approach is based on provisioning a new VM on a node which currently (immediately prior to provisioning) has the maximum resource consumption. Another example approach is based on provisioning a new VM on a node which currently has small-sized VMs in terms of resource requirements. In yet another embodiment, the approach is based on provisioning a new VM on a node located in a geographical area having low power tariffs.