Intel Corporation

Splunk Inc.

Cisco Technology, Inc.

At&t Intellectual Property I, L.p.

Mohan Rajagopalan

Ram Sriharsha

John C Reed

Amir Malekpour

Chinmay Madhav Kulkarni

Bratin Saha

Shoumeng Yan

Lin Ma

Hu Tiger Chen

Ying Gao

Avi Mendelson

Peinan Zhang

Xiaocheng Zhou

Anwar Ghuloum

Jesse Z Fang

Zhou Xiaocheng

Joshua B Fryman

Ali-Reza Adl-Tabatabai

Milind Baburao Girkar

Stephen Junkins

Matti A Hiltunen

David M Putzolu

Robert Geva

Richard Dale Schlichting

Richard Lewis Hudson

Yang Ni

Sai Luo

Adam Welc

Trevor Jim

Matthew Hammer

Rajiv Deodhar

Clark F Nelson

Leaf Petersen

Ravi Narayanaswamy

Patrick Xi

Gao Ying

Tiger Chen

Hu Chen

A computer-implemented method is disclosed that includes operations of receiving a request to store a representation of a machine learning (ML) model in a non-transitory computer-readable medium, validating the representation of the ML model, and storing the representation of the ML model. The representation of the ML model may be a computation graph, and further, the computation graph may be a format for representing ML models through a defined common set of operators and a common file format. Validating the representation of the ML model may include verifying an input schema of the representation of the ML model is compatible with an input schema of the ML model. The method may include further operations of deserializing the representation of the ML model, wherein a successful deserialization indicates that the representation of the ML model is valid and secure.

Systems and methods for securing untrusted code

A method for deployment of machine-learning based operators within a query is described. For this embodiment, a sequence of operators associated with a query is identified, which includes at least a first operator and at least a second operator. The second operator is configured to perform operations, in accordance with a machine learning (ML) component, on data received as input from execution of the first operator. Schemas associated with the machine learning component is retrieved along with schemas associated with other operators within the sequence. Compatibility between at least an output schema associated with the first operator and an input schema associated with the second operator associated with the ML component is determined. Thereafter, a portion of the sequence of operators including at least the second operator and another operator of the sequence of operators successive to the second operator may be stored within a data store for subsequent use.

Systems and methods for auto-deployment of a machine learning component within a pipelined search query

Disclosed herein is a method that supports queries deploying operators based on multiple programming languages at least through determining schema compatibility between neighboring operators within a query. Upon receipt of a query, a sequence of operators of the query is identified, where the sequence of operators includes at least two neighboring operators including a first operator and a second operator representing a machine learning model. By determining schema compatibility between at least the first and second operators, the method either alerts a user to schema incompatibility before attempting to execute the query or determine that the schemas are compatible such that the query may be executed without the occurrence of errors due to schema incompatibility between neighboring operators. Advantageously, the method enables the integration of a machine learning model into the query while still ensuring schema compatibility.

Determination of schema compatibility between neighboring operators within a search query statement

A computer-implemented method is disclosed that includes operations of parsing a query comprised of a sequence of operators to detect each operator of the sequence of operators, where the sequence of operators includes a machine learning (ML) operator representing a trained ML model. Additionally, a schema of the ML operator is determined through metadata. A filter or a projection is generated based on the schema of the ML operator, where the filter or projection is configured to reduce an amount of data retrieved upon application of the filter of the projection to an operator of the sequence of operators comprising the query. The schema of the ML operator indicates a schema of input data to be provided to the ML operator and a schema of output data to be provided by the ML operator following processing.

Systems and methods for decoupling search processing language and machine learning analytics from storage of accessed data

A computer-implemented method for integration of machine learning components within a pipelined search query to generate a visualization is described. Herein, an interface is provided for receipt of pipelined code into a web-based programming application. The pipelined code features a series of operators configured to perform one or more tasks based on collective operations by the series of operators, wherein a first operator of the series of operators is to receive input data from a selected data source and each remaining operator of the series of operators to receive input based on an output from a preceding operator of the remaining operators. The task(s) performed by the pipelined code generate results including visualizations. The visualization is rendered in a manner that allows the pipelined code to be scrolled to display the pipelined code or the visualization.

Systems and methods for integration of machine learning components within a pipelined search query to generate a graphic visualization

A computer-implemented method is disclosed including operations of receiving a request to store a representation of a machine learning model in a non-transitory computer-readable medium, validating the representation of the machine learning model, storing the representation of the machine learning model, receiving a query from a web-based programming application, the query including a sequence of operators, parsing the query to detect and identify each operator within the sequence of operators, converting the query to directed acyclic graph (DAG) and providing the DAG to a distributed processing engine configured to execute the DAG. The computer-implemented method includes further operations of, prior to converting the query to the DAG, altering the query to improve efficiency of execution of the DAG. Altering the query may include at least one of consolidating at least two operators, applying a filter operation to an operator, or applying a projection to the operator.

Systems and methods for a unified analytics platform

According to one embodiment, a method that supports queries deploying operators based on multiple programming languages is described. A sequence of operators associated with a query is identified, where the sequence of operators includes at least two neighboring operators including a first operator based on a first programming language and a second operator based on a second programming language that is different from the first programming language. Thereafter, a schema associated with the first operator and a schema associated with the second operator is determined along with the compatibility between the schema of the first operator and the schema of the second operator. A query error message is generated in response to incompatibility between the first operator schema and the second operator schema. Compatibility is determined when an output generated by execution of the first operator provides machine data needed as input for execution of the second operator.

Systems and methods for integration of multiple programming languages within a pipelined search query

A computer-implemented method is disclosed that includes operations of receiving a query to be executed, the query including an indication of a data source at which input data is be to obtained, wherein the query is to be executed on the input data, determining a schema of the input data, determining fields of the input data that are required for execution of the query by analyzing a sequence of operators forming the query, determining one or more alterations to the query to improve efficiency of the execution of the query based on the fields of input data required for the execution, and generating an altered query be altering the query in accordance with the one or more alterations. The method may further include converting the query to a directed acyclic graph (DAG) and providing the DAG to a distributed processing engine configured to execute the DAG.

Embodiments of the invention provide a programming model for CPU-GPU platforms. In particular, embodiments of the invention provide a uniform programming model for both integrated and discrete devices. The model also works uniformly for multiple GPU cards and hybrid GPU systems (discrete and integrated). This allows software vendors to write a single application stack and target it to all the different platforms. Additionally, embodiments of the invention provide a shared memory model between the CPU and GPU. Instead of sharing the entire virtual address space, only a part of the virtual address space needs to be shared. This allows efficient implementation in both discrete and integrated settings.

Shared virtual memory

Embodiments of the invention provide language support for CPU-GPU platforms. In one embodiment, code can be flexibly executed on both the CPU and GPU. CPU code can offload a kernel to the GPU. That kernel may in turn call preexisting libraries on the CPU, or make other calls into CPU functions. This allows an application to be built without requiring the entire call chain to be recompiled. Additionally, in one embodiment data may be shared seamlessly between CPU and GPU. This includes sharing objects that may have virtual functions. Embodiments thus ensure the right virtual function gets invoked on the CPU or the GPU if a virtual function is called by either the CPU or GPU.

Language level support for shared virtual memory

Methods, systems, and mediums are described for scheduling data parallel tasks onto multiple thread execution units of processing system. Embodiments of a lock-free queue structure and methods of operation are described to implement a method for scheduling fine-grained data-parallel tasks for execution in a computing system. The work of one of a plurality of worker threads is wait-free with respect to the other worker threads. Each node of the queue holds a reference to a task that may be concurrently performed by multiple thread execution units, but each on a different subset of data. Various embodiments relate to software-based scheduling of data-parallel tasks on a multi-threaded computing platform that does not perform such scheduling in hardware. Other embodiments are also described and claimed.

Systems and methods for data-parallel processing

A computer system may comprise a computer platform and input-output devices. The computer platform may include a plurality of heterogeneous processors comprising a central processing unit (CPU) and a graphics processing unit) GPU, for example. The GPU may be coupled to a GPU compiler and a GPU linker/loader and the CPU may be coupled to a CPU compiler and a CPU linker/loader. The user may create a shared object in an object oriented language and the shared object may include virtual functions. The shared object may be fine grain partitioned between the heterogeneous processors. The GPU compiler may allocate the shared object to the CPU and may create a first and a second enabling path to allow the GPU to invoke virtual functions of the shared object. Thus, the shared object that may include virtual functions may be shared seamlessly between the CPU and the GPU.

Two way communication support for heterogenous processors of a computer platform

A method for concurrent management of adaptive programs is disclosed wherein changes in a set of modifiable references are initially identified. A list of uses of the changed references is next computed using records made in structures of the references. The list is next inserted into an elimination queue. Comparison is next made of each of the uses to the other uses to determine independence or dependence thereon. Determined dependent uses are eliminated and the preceding steps are repeated for all determined independent uses until all dependencies have been eliminated.

Concurrent management of adaptive programs

Growing community of inventors

Mountain View, CA, United States of America

Mohan Rajagopalan