Distributed neuromorphic processing performance accountability

Patent metadata

Field	Value
Publication number	US-10338968-B2
Application number	US-201816039863-A
Country	US
Kind code	B2
Filing date	Jul 19, 2018
Priority date	Feb 5, 2016
Publication date	Jul 2, 2019
Grant date	Jul 2, 2019

Abstract

An apparatus includes a processor to: receive a request to repeat an earlier performance of a first job flow described in a job flow definition; analyze the job flow definition to determine whether the first job flow uses a neural network; in response to a determination that the first job flow uses a neural network, analyze an object associated with the first job flow to determine whether the neural network was trained using training data from a second job flow that does not use a neural network; and in response to a determination that such training data was so used, repeat the earlier performance of the first job flow, perform the second job flow with the same input data values as used in the repeated performance of the first job flow, and analyze corresponding output data values of both performances to determine a degree of accuracy of the neural network.

First claim

The invention claimed is:

1. An apparatus comprising a processor and a storage to store instructions that, when executed by the processor, cause the processor to perform operations comprising:
receive, at a portal, and from a remote device via a network, a request to repeat an earlier performance, described in a first instance log of multiple instance logs stored in one or more federated areas, of a first job flow defined in a first job flow definition of multiple job flow definitions stored in the one or more federated areas, or to provide objects to the remote device to enable the remote device to repeat the earlier performance, wherein:
the portal is provided on the network to control access by the remote device to the one or more federated areas via the network;
the one or more federated areas are maintained within one or more storage devices to store at least the multiple job flow definitions and the multiple instance logs; and
the request specifies a first instance log identifier of the first instance log;

use the first instance log identifier to retrieve the first instance log from among the multiple instance logs stored in the one or more federated areas, wherein the first instance log comprises a first job flow identifier of the first job flow definition, a task routine identifier for each task routine used to perform a task specified in the first job flow definition, and a data object identifier for each data object associated with the earlier performance of the first job flow;
analyze the first job flow definition to determine whether performances of the first job flow comprise use of a neural network;
in response to a determination that performances of the first job flow do comprise use of the neural network, analyze an object associated with the first job flow to determine whether the neural network was trained to perform an analytical function using a training data set derived from at least one performance of a second job flow defined by a second job flow definition stored in the one or more federated areas, wherein:
the object associated with the first job flow comprises at least one of the first job flow definition, the first instance log, or a task routine executed during the earlier performance of the first job flow; and
performances of the second job flow comprise performances of the analytical function in a manner that does not use any neural network; and

in response to the request comprising a request to repeat the earlier performance, in response to a determination that performances of the first job flow do comprise use of the neural network, and in response to a determination that the neural network was trained using the training data set derived from at least one performance of the second job flow, the processor is caused to perform operations comprising:
repeat the earlier performance of the first job flow with one or more data sets associated with the earlier performance of the first job flow, wherein the repetition of the earlier performance of the first job flow comprises execution, by the processor, of each task routine identified by a task routine identifier in the first instance log;
perform the second job flow with the one or more data sets associated with the earlier performance of the first job flow, wherein the performance of the second job flow comprises execution, by the processor, of a most recent version of a task routine to perform each task identified by a flow task identifier in the second job flow definition;
analyze an output of the repetition of the earlier performance of the first job flow relative to a corresponding output of the performance of the second job flow to determine a degree of accuracy of the first job flow in performing the analytical function relative to a predetermined threshold of accuracy to determine whether the second job flow is to be used in place of the first job flow to perform the analytical function; and
transmit at least the output of the repetition of the earlier performance of the first job flow and an indication of the degree of accuracy or the results of the comparison to the requesting device.

2. The apparatus of claim 1 , wherein the processor is caused to perform operations comprising:
use the first job flow identifier within the first instance log to retrieve the first job flow definition from the one or more federated areas;
for each task routine executed during the earlier performance of the first job flow to perform a corresponding task specified in the first job flow definition, use the corresponding task routine identifier within the first instance log to retrieve the task routine from the one or more federated areas; and
for each data object employed as an input to the earlier performance of the first job flow, use the corresponding data object identifier within the first instance log to retrieve the data object from the one or more federated areas.

3. The apparatus of claim 2 , wherein the processor is caused to, in response to the request comprising a request to provide objects to the remote device to enable the remote device to repeat the earlier performance, transmit, to the requesting device, at least the first job flow definition, each task routine executed during the earlier performance of the first job flow, and each data object employed as an input to the earlier performance of the first job flow.

4. The apparatus of claim 2 , wherein the processor is caused to, in response to the request comprising a request to repeat the earlier performance of the first job flow, perform operations comprising:
for each data object generated during the earlier performance of the first job flow, use the corresponding data object identifier within the first instance log to retrieve the data object from the one or more federated areas, and analyze a corresponding data object generated during the repetition of the earlier performance relative to a corresponding data object generated during the earlier performance; and
transmit, to the requesting device, an indication of results of the analysis of at least one data object generated during the repetition of the earlier performance relative to at least one corresponding data object generated during the earlier performance.

5. The apparatus of claim 2 , wherein the processor is caused, in response to the determination that performances of the first job flow do comprise use of a neural network, and in response to a determination that the neural network was trained using the training data set derived from at least one performance of the second job flow, to perform operations comprising:
use a second job flow identifier provided within the selected object associated with the first job flow to retrieve the second job flow definition from the one or more federated areas;
for each task specified by a corresponding flow task identifier in the second job flow definition, use the corresponding flow task identifier to retrieve a most recent version of a corresponding task routine from the one or more federated areas; and
in response to the request comprising a request to provide objects to the remote device to enable the remote device to repeat the earlier performance, transmit, to the requesting device, at least the second job flow definition and the most recent version of task routine corresponding to each task specified in the second job flow definition.

6. The apparatus of claim 2 , wherein the processor is caused, in response to the determination that performances of the first job flow do comprise use of a neural network, and in response to a determination that the neural network was trained using the training data set derived from at least one performance of the second job flow, to perform operations comprising:
use a second instance log identifier provided within

Assignees

• Sas Inst Inc

Inventors

• Bequet Henry Gabriel Victor
• Chen Iii Huina
• Du Juan

Classifications

• G06F9/5027 Primary

  the resource being a machine, e.g. CPUs, Servers, Terminals · CPC title

• H04L67/10 Primary

  in which an application is distributed across nodes in the network (software deployment G06F8/60; multiprogramming arrangements G06F9/46) · CPC title

• G06F18/217

  Validation; Performance evaluation; Active pattern learning techniques · CPC title

• G06F9/4843

  by program, e.g. task dispatcher, supervisor, operating system · CPC title

• G06N3/063

  using electronic means · CPC title