Unlock instant, AI-driven research and patent intelligence for your innovation.

A search method, device and electronic equipment for distributed model parameters

A technology of model parameters and search methods, applied in the field of machine learning, can solve problems such as low efficiency of parameter tuning, and achieve the effect of improving tuning efficiency

Active Publication Date: 2021-05-18
HANGZHOU FRAUDMETRIX TECH CO LTD
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] In view of this, the present invention provides a search method, device and electronic equipment for distributed model parameters to solve the problem of low parameter tuning efficiency when the parameter space dimension is high

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A search method, device and electronic equipment for distributed model parameters
  • A search method, device and electronic equipment for distributed model parameters
  • A search method, device and electronic equipment for distributed model parameters

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0033] figure 1 It is a flowchart of a search method for distributed model parameters according to an embodiment of the present application.

[0034] refer to figure 1 As shown, the search method provided in this embodiment is used to search the distributed model parameters, so as to find the best parameter combination of the corresponding model, the distributed model parameters are parameters or hyperparameters obtained during model training, The search method specifically includes the following steps:

[0035] S1. Perform verification calculation for each gridded space.

[0036] That is, for each specific grid space, the verification calculation is performed using the pre-acquired verification sample set, so as to obtain the cross-validation mean value of the evaluation index corresponding to each grid space.

[0037] The grid space here refers to the parameter combination or hyperparameter combination obtained by random combination of distributed model parameters obtaine...

Embodiment 2

[0051] image 3 It is a block diagram of a device for searching distributed model parameters according to an embodiment of the present application.

[0052] refer to image 3 As shown, the search device provided in this embodiment is used to search the distributed model parameters, so as to find out the best parameter combination of the corresponding model, the distributed model parameters are parameters or hyperparameters obtained during model training, The search device specifically includes a grid search module 10 , a space narrowing module 20 and a parameter tuning module 30 .

[0053] A grid search module is used to perform validation calculations for each gridded space.

[0054] That is, for each specific grid space, the verification calculation is performed using the pre-acquired verification sample set, so as to obtain the cross-validation mean value of the evaluation index corresponding to each grid space. There are multiple grid search modules here.

[0055] The ...

Embodiment 3

[0068] This embodiment provides an electronic device, which includes at least one processor and a corresponding memory, and the processor and the memory are connected through a corresponding data bus. The memory is used to store computer programs or instructions, and the processor is used to execute computer programs or instructions in order to:

[0069] For each grid space, use the pre-acquired verification sample set to perform verification calculations, and obtain the cross-validation mean value of the evaluation index corresponding to each grid space, and the grid space is the training distributed model. A random combination between the obtained parameters; select the current optimal cross-validation mean value from the obtained multiple cross-validation mean values, and select the most likely local search space according to the current optimal cross-validation mean value; The pre-selected optimal sampling points are subjected to Bayesian optimization calculation in the lo...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

Embodiments of the present invention provide a search method, device, and electronic equipment for distributed model parameters. The method and device are specifically for each gridded space, using pre-acquired verification sample sets to perform verification calculations to obtain The cross-validation mean value of the evaluation index corresponding to the grid space; select the current optimal cross-validation mean value from the obtained multiple cross-validation mean values, and select the most likely local search space according to the current optimal cross-validation mean value; The selected best sampling points are subjected to Bayesian optimization calculation in the local search space, and the optimal parameter combination in the local search space is obtained. Since the solution first narrows the search space based on the corresponding rules, and then performs parameter search in the narrowed search space, the efficiency of parameter tuning is improved when the dimension of the parameter space is high.

Description

technical field [0001] The invention relates to the technical field of machine learning, in particular to a search method, device and electronic equipment for distributed model parameters. Background technique [0002] In the application of machine learning, many parameters are generated through training, so that these parameters can be used to construct corresponding models. In order to make the final function have the best effect, it is necessary to find out the optimal parameter or parameter combination, and substitute it into the corresponding function to finally construct the model, that is, parameter tuning. When there are few parameters or parameter combinations, the purpose of tuning can be met by random search; when there are many parameters or parameter combinations, that is, the dimension of the parameter space is high, random search needs to randomly select a large number of points in a larger parameter space , resulting in high time cost and low tuning efficien...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06F16/27G06K9/62
Inventor 吴浩然顾全
Owner HANGZHOU FRAUDMETRIX TECH CO LTD