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

Energy-efficient and storage-efficient training of neural networks

An artificial neural network and training data technology, applied in the field of neural network training, can solve problems such as consumption and energy consumption, and achieve the effect of reducing storage consumption, saving computing time and energy consumption

Pending Publication Date: 2022-05-27
ROBERT BOSCH GMBH
View PDF1 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The training is typically very computationally expensive and accordingly consumes a lot of energy

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
  • Energy-efficient and storage-efficient training of neural networks
  • Energy-efficient and storage-efficient training of neural networks
  • Energy-efficient and storage-efficient training of neural networks

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0043] figure 1 is a schematic flow diagram of one embodiment of a method 100 for training KNN1. In step 105, KNN 1 is optionally selected, which KNN is constructed as an image classifier.

[0044] At step 110, the trainable parameters 12 of KNN 1 are initialized. According to block 111 , the values ​​for this initialization can be derived, for example, from a sequence of numbers, which the deterministic algorithm 16 provides based on the starting configuration 16 a . According to block 111a, the sequence of numbers may in particular be, for example, a pseudo-random sequence of numbers.

[0045] In step 120, training data 11a is provided. These training data are labeled with the nominal output 13a to which KNN 1 should map the training data 11a, respectively.

[0046] Training data 11a is fed to KNN 1 in step 130 and mapped to output 13 by KNN 1 . In a step 140 , the correspondence of these outputs 13 with the learning outputs 13 a is evaluated on the basis of a predefine...

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 training an artificial neural network KNN whose own behavior is characterized by trainable parameters includes: initializing parameters; providing training data marked with rated outputs, wherein the KNN respectively maps the training data to the rated outputs; training data is transmitted to the KNN and mapped to output by the KNN; evaluating the consistency of the output and the learning output according to a preset cost function; selecting at least a first subset of parameters to be trained and a second subset of parameters to be maintained from the parameter set according to a predetermined criterion; parameters to be trained are optimized with the following objectives: further processing of training data by KNN is expected to result in better evaluation through a cost function; the parameter to be maintained is retained on its initialization value or on a value that has been obtained during optimization, respectively.

Description

technical field [0001] The present invention relates to the training of neural networks, which can be used, for example, as image classifiers. Background technique [0002] The artificial neural network KNN maps inputs, such as images, to outputs relevant for the respective application by means of a processing chain, which is characterized by a large number of parameters and can be organized, for example, in the form of layers. For example, an image classifier provides as output an assignment to one or more classes of a predefined classification for an input image. The KNN is trained in such a way that training data are fed to the KNN and the parameters of the processing chain are optimized such that the output provided corresponds as well as possible to the previously known nominal output belonging to the respective training data. [0003] The training is typically very computationally intensive and accordingly consumes a lot of energy. In order to reduce the computationa...

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): G06N3/08G06N3/04G06K9/62G06V10/764G06V10/82
CPCG06N3/084G06N3/045G06F18/24G06N3/08G06N3/04
Inventor A·P·孔杜拉凯J·E·M·梅纳特P·维默
Owner ROBERT BOSCH GMBH