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

Method and device for transfer learning between modified tasks

A transfer learning and machine learning technology, applied in the field of machine-readable storage media, can solve the problem of suboptimal hyperparameter configuration

Pending Publication Date: 2022-02-22
ROBERT BOSCH GMBH
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The hyperparameter configurations used so far may be suboptimal if, for example, an expansion of the search space is performed or by adding new hyperparameters of the machine learning algorithm

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 and device for transfer learning between modified tasks
  • Method and device for transfer learning between modified tasks
  • Method and device for transfer learning between modified tasks

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0039] figure 1 A flowchart of a method according to an embodiment of the invention is schematically shown.

[0040] By way of example, this embodiment explains how the invention can be applied to a machine learning algorithm with a plurality of hyperparameters. Hyperparameters can be: learning rate, number of neurons in a layer, number of layers, batch size (English batchsize). Additionally, a classification hyperparameter can also be given, such as a Boolean variable that indicates whether augmented training data is used. The machine learning algorithm may be gradient descent for optimizing the weights of a neural network with respect to a cost function.

[0041] The method starts at step S11. This step consists in providing the search space by specifying the range of values ​​for the hyperparameters . The value range of the learning rate can be specified, for example, such that the value range is from 0.0001 to 0.99. In step S11, the cost function of the machine lear...

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 for the transfer learning of hyperparameters of a machine learning algorithm. The method includes providing a current search space and a previous search space. A reduced search space is then created and candidate configurations are drawn repeatedly at random from the reduced search space and from the current search space, and the machine learning algorithm, parameterized in each case with the candidate configurations, is applied. A Tree Parzen Estimator (TPE) is then created as a function of the candidate solutions and the results of the machine learning algorithm parameterized with the candidate configurations, and the drawing of further candidate configurations from the current search space using the TPE is repeated multiple times, the TPE being updated upon each drawing.

Description

technical field [0001] The present invention relates to a method for transfer learning between modified tasks for a machine learning method as well as a computer program and a machine-readable storage medium. Background technique [0002] Machine learning algorithms often require expensive hyperparameter optimization (English H yper p parameters O optimization, HPO), so that only in this way can these algorithms achieve their optimal performance for a given task setting. Unfortunately, the optimal hyperparameter configuration may change each time, for example, when a developer makes a (small) adaptation to the algorithm or the search space, whereby this hyperparameter configuration has to be reset. Because machine learning methods are known to be very sensitive about their hyperparameters. Thus, it is common practice to completely restart the HPO after the just mentioned adaptation. Since all knowledge acquired in previous development steps is not taken into account in ...

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): G06N20/00G06N3/04G06N3/08
CPCG06N20/00G06N3/04G06N3/08G06N3/082G06N20/10G06V10/82G06V20/56G06V40/10G06V40/20G06N5/01G06N7/01G06F18/214
Inventor D·斯托尔D·瓦格纳F·胡特尔J·弗兰克S·赛尔格
Owner ROBERT BOSCH GMBH