US 2021/0200575 A1Published application

Self-Optimizing Computation Graphs

AssigneePaypal, Inc.

InventorDavid Williams

Life of this patent

Estimating the term…

Loading patent…

Abstract

A method includes receiving code of an application, the code structured as a plurality of instructions in a computation graph that corresponds to operational logic of the application. The method also includes processing the code according to an iterative learning process. The iterative learning process includes determining whether to adjust an exploration rate associated with the iterative learning process based on a state of a computing environment. Additionally, the process includes executing the plurality of instructions of the computation graph according to an execution policy that indicates certain instructions to be executed in parallel. The process also includes determining an execution time for executing the plurality of instructions of the computation graph according to the execution policy and based on the execution time and the exploration rate, adjusting the execution policy to reduce the execution time in a subsequent iteration.

Claims20

Dependent claims

A system, comprising:

one or more hardware processors;
and
at least one memory storing computer-executable instructions, that in response to execution by the one or more hardware processors, causes the system to perform operations comprising:
receiving code of an application, the code structured as a plurality of instructions in a computation graph that corresponds to operational logic of the application;
processing the code according to an iterative learning process, each iteration of the iterative learning process comprising:
determining whether to adjust an exploration rate associated with the iterative learning process based on a state of a computing environment in which the code is being processed;
executing the plurality of instructions of the computation graph according to an execution policy, the execution policy indicating a first subset of the plurality of instructions to be executed in parallel with a second subset of the plurality of instructions;
determining an execution time for executing the plurality of instructions of the computation graph according to the execution policy;
based on the execution time and the exploration rate, adjusting the execution policy to reduce the execution time in a subsequent iteration;
and
decreasing the exploration rate;
and
determining, based on the iteratively processing the code, a final execution policy.

10 more elements

A non-transitory computer readable medium storing computer-executable instructions that in response to execution by one or more hardware processors, causes a system to perform operations comprising:

receiving code of an application, the code structured as a plurality of instructions in a computation graph that corresponds to operational logic of the application;
processing the code according to an iterative learning process, each iteration of the iterative learning process comprising:
executing the plurality of instructions of the computation graph according to an execution policy, the execution policy indicating a first subset of the plurality of instructions to be executed in parallel with a second subset of the plurality of instructions;
determining an execution time for executing the plurality of instructions of the computation graph;
based on the execution time and an exploration rate associated with the iterative learning process, adjusting the execution policy to reduce the execution time in a subsequent iteration;
and
decreasing the exploration rate;
and
determining, based on the iteratively processing the code, a final execution policy.

6 more elements

Dependencies are inferred from the claim text.

Facts

Application: 16/732,029Filed 31 Dec 2019
Published: 1 Jul 2021
Term ends: Not estimatedThe source records don’t support an estimate.

Classification7 codes

G06F 8/45
Compilation: exploiting coarse grain parallelism in compilation, i.e. parallelism between groups of instructions
G06F 9/463
Saving or restoring of program or task context: program control block organisation
G06N 5/01
Computing arrangements using knowledge-based models: dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
G06N 5/022
Knowledge representation; Symbolic representation: knowledge engineering; Knowledge acquisition