Method and device for creating a data-based function model

Patent metadata

Field	Value
Publication number	US-9709967-B2
Application number	US-201414247104-A
Country	US
Kind code	B2
Filing date	Apr 7, 2014
Priority date	Apr 10, 2013
Publication date	Jul 18, 2017
Grant date	Jul 18, 2017

Abstract

A method for generating a data-based function model includes: providing a first data-based partial model ascertained from a first training data record; providing at least one additional training data record; and performing the following steps for the at least one additional training data record: ascertaining a difference training data record having training data which correspond to the differences between the output values of the relevant additional training data record and the function value of the sum of the partial function values (f first _ partial _ model (x) f second _ partial _ model (x)) of the first data-based partial model and previously ascertained data-based partial model(s) at each of the measuring points of the relevant training data record; ascertaining an additional data-based partial model from the difference training data record; and forming a sum (f(x)) from the first and the additional data-based partial models.

First claim

What is claimed is:

1. A method for calculating a function value of a data-based function model in a control unit which has an arithmetic unit and a separate model calculation unit, wherein the data-based function model has multiple partial models, comprising:
providing input data to the model calculation unit;
transmitting a calculation address, which indicates an address of first configuration data in a memory unit, to a direct memory access (DMA) unit;
retrieving the first configuration data, which contain parameters and node data for calculating a first partial model, from the memory unit, and transmitting the first configuration data to the model calculation unit;
ascertaining a first partial function value of the first partial model based on the input data;
storing the first partial function value in the model calculation unit; and
repeating the following steps until a predefined stop condition is satisfied:
retrieving additional configuration data, which contain parameters and node data for calculating an additional partial model, from the memory unit, and transmitting the relevant additional configuration data to the model calculation unit;
ascertaining an additional partial function value of the additional partial model based on the input data; and
adding the additional partial function value to the value stored in the model calculation unit.

2. The method as recited in claim 1 , wherein the additional configuration data are retrieved from a memory area which is determined by applying a predefined address offset to the calculation address.

3. The method as recited in claim 2 , wherein the stop condition is specified by the retrieved additional configuration data.

4. The method as recited in claim 3 , wherein the stop condition specifies that the arithmetic unit indicates the value stored in the model calculation unit as a function value of a data-based function model.

5. The method as recited in claim 4 , wherein the retrieval of the additional configuration data from the memory unit, the transmission of the relevant additional configuration data to the model calculation unit, and the starting of the ascertainment of the additional partial function value of the additional partial model are carried out by the DMA unit.

6. A control unit for calculating a function value of a data-based function model in the form of a Gaussian process model which has multiple partial models, comprising:
a memory unit to store first configuration data which contain parameters and node data for calculating a first partial model, and additional configuration data which contain parameters and node data for calculating at least one additional partial model;
a direct memory access (DMA) unit;
an arithmetic unit to transmit a calculation address, which indicates an address of the first configuration data in the memory unit, to the DMA unit; and
a separate model calculation unit;
wherein the DMA unit is configured to:
retrieve the first configuration data from the memory unit and transmit the first configuration data to the model calculation unit; and
retrieve the additional configuration data from the memory unit and transmit the additional configuration data to the model calculation unit; and

wherein the model calculation unit is configured to:
ascertain a first partial function value of the first partial model based on the input data;
store the first partial function value in the model calculation unit;
ascertain an additional partial function value of the at least one additional partial model based on the input data; and
add the additional partial function value to the value stored in the model calculation unit.

7. A non-transitory, computer-readable data storage medium storing a computer program having program codes which, when executed on a computer, performs a method for calculating a function value of a data-based function model in a control unit which has an arithmetic unit and a separate model calculation unit, wherein the data-based function model has multiple partial models, the method comprising:
providing input data to the model calculation unit;
transmitting a calculation address, which indicates an address of first configuration data in a memory unit, to a direct memory access (DMA) unit;
retrieving the first configuration data, which contain parameters and node data for calculating a first partial model, from the memory unit, and transmitting the first configuration data to the model calculation unit;
ascertaining a first partial function value of the first partial model based on the input data;
storing the first partial function value in the model calculation unit; and
repeating the following steps until a predefined stop condition is satisfied:
retrieving additional configuration data, which contain parameters and node data for calculating an additional partial model, from the memory unit, and transmitting the relevant additional configuration data to the model calculation unit;
ascertaining an additional partial function value of the additional partial model based on the input data; and
adding the additional partial function value to the value stored in the model calculation unit.

Assignees

• Bosch Gmbh Robert

Inventors

• Markert Heiner
• Diener Rene
• Streichert Felix
• Guntoro Andre
• Hanselmann Michael

Classifications

• G06N99/005

  Physics · mapped topic

• G05B13/04 Primary

  involving the use of models or simulators · CPC title

• G06N20/00 Primary

  Machine learning · CPC title