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

Method, equipment, and computer program for predicting learning curve

A learning curve, machine learning technology, applied in neural learning methods, computer systems based on knowledge-based models, computing, etc., can solve problems such as unpredictability and inefficient attempts, and achieve high performance and save computer resources.

Pending Publication Date: 2020-12-01
ROBERT BOSCH GMBH
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the attempt is extremely inefficient, since all calculations have to be performed each time for this, and it is not possible to predict whether the adapted hyperparameters of the training method subsequently lead to a better learning curve

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
  • Method, equipment, and computer program for predicting learning curve
  • Method, equipment, and computer program for predicting learning curve
  • Method, equipment, and computer program for predicting learning curve

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0039] figure 1 A first machine learning system ( 10 ) trained by means of a device ( 11 ) is schematically shown. The device (11) obtains training input variables which are processed into output variables by means of a first machine learning system (10). The training input variables can be as exemplarily in figure 1 As shown in the image. However, other signals, such as audio signals, are also conceivable. After determining the output variable of the first machine learning system (10), the determined output is determined by means of a cost function (English loss function (loss function)) from the determined output variable and the training output variable provided for the training input variable The difference between the variable and the training output variable. The cost function then outputs the value . If the first machine learning system ( 10 ) is a deep neural network, the cost function may eg be squared error.

[0040] the value It can then be used to adapt t...

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

A method, equipment, and a computer program for predicting a learning curve. The invention relates to a method (30) for predicting a learning curve of an iterative training method. The method (30) comprises a step of obtaining a hyper-parameter theta of the training method and at least one preceding value of a cost function and a step of predicting a learning curve from the hyper-parameter theta and from the at least one preceding value of the learning curve by means of a second machine learning system (12). Furthermore, the invention relates to a computer program and to equipment for carryingout the method (30), and to a machine-readable storage element on which the computer program is stored.

Description

technical field [0001] The present invention relates to a method for predicting a learning curve in an iterative training method of a machine learning system. Likewise, the invention relates to a device and a computer program, respectively configured to carry out the method. Background technique [0002] Authors T. Elsken, J. Metzen, and F. Hutter give an overview of methods for optimizing the architecture of machine learning systems in their publication "Neural architecturesearch: A survey." (arXiv preprint arXiv:1808.05377 (2018)) . [0003] The authors A. Klein, S. Falkner, J. T. Springenberg, and F. Hutter in their publication "Learning curve prediction with Bayesian neural networks" (International Conference on Learning Representations (ICLR'17)) disclose a machine learning system for predicting approach with a learning curve. [0004] Training of machine learning systems is very computationally intensive and thus extremely time consuming even on high performance com...

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 Applications(China)
IPC IPC(8): G06N20/00G06N20/10G06N3/08G06N3/04G06N5/00
CPCG06N20/00G06N20/10G06N3/08G06N5/01G06N3/047G06N3/044G06N3/045G06N20/20G06N7/01
Inventor A.克莱因F.胡特M.加贾尼S.法尔克纳
Owner ROBERT BOSCH GMBH