EP3931760A1

EP3931760A1 - Training neural networks for efficient implementation on hardware

Info

Publication number: EP3931760A1
Application number: EP20705699.5A
Authority: EP
Inventors: Falk Heuer; Philip Lenz; Alexander Lengsfeld
Original assignee: Robert Bosch GmbH
Current assignee: Robert Bosch GmbH
Priority date: 2019-03-01
Filing date: 2020-02-17
Publication date: 2022-01-05
Also published as: WO2020178009A1; CN113454655A; US20220101088A1; DE102019202816A1

Abstract

A method (100) for training an artificial neural network, ANN (1), which comprises a plurality of neurons (2), comprising the following steps: • a measure of the quality (11) that the ANN (1) has achieved overall within a past time period is determined (110); • one or more neurons (2) are evaluated on the basis of a measure of their respective quantitative contributions (21) to the determined quality (11) (120); • actions (22), by means of which the evaluated neurons (2) are respectively trained in the further course of the training, and/or weights (23) of said neurons (2) in the ANN (1) are defined on the basis of the evaluations (120a) of the neurons (2) (130). The method (200) according to claim 11, wherein a computing unit (4) having hardware resources for a specified number of neurons (2), layers (3a, 3b) of neurons (2) and/or connections (25) between neurons (2) is selected (205a), and wherein a model (1a) of the ANN (1) having a number of neurons (2), layers (3a, 3b) of neurons (2) and/or connections (25) between neurons (2) that exceeds the specified number is selected (205b).

Description

description

Title:

Training of neural networks for efficient implementation on hardware

The present invention relates to the training of neural networks with the aim of being able to implement these neural networks efficiently on hardware, for example for use on board vehicles.

State of the art

An artificial neural network, ANN, comprises an input layer, several processing layers and an output layer. Input variables are read into the ANN at the input layer and on their way through the processing layers to the output layer using a

Processing chain processed, which is usually parameterized. When training the ANN, those values are determined for the parameters of the processing chain with which the processing chain can obtain a set of learning values for the

Input variables optimally based on an associated set of learning values for the

Maps output variables.

The strength of KNN lies in the fact that they can process very high-dimensional data, such as high-resolution images, massively in parallel. The price for this parallel processing is a high hardware expenditure for the implementation of an ANN. Graphics processors (GPUs) with a large amount of memory are typically required. Based on the knowledge that a large part of the neurons in a deep ANN make little or no contribution to the end result delivered by the ANN, US Pat. No. 5,636,326 A discloses the weights of connections between neurons in the fully trained ANN in a reading process (“pruning ") to undergo. This allows the number of Connections and neurons are greatly reduced without much loss of accuracy.

Disclosure of the invention

In the context of the invention, a method for training an artificial neural network, ANN, was developed. The aim is to be able to implement the ANN efficiently on hardware. In this context, “efficient” can be understood, for example, to mean that the ANN gets by with a limited configuration of hardware resources. “Efficient” can also be understood, for example, that the

Network architecture and / or the neurons in one or more layers of the ANN are optimally used and / or fully utilized. The exact definition of "efficient" results from the specific application in which the ANN is used.

In the method, a measure of the quality that the ANN and / or a sub-area of the ANN has achieved overall within a previous period is determined at any point in time during the training, which otherwise takes place in any known manner. The quality can include, for example, training progress, utilization of the neurons of a layer or another sub-area of the ANN, utilization of the neurons of the ANN as a whole, as well as any, for example weighted, combinations thereof. The exact definition of “quality” thus also results from the specific application.

Thus, the measure for the quality, for example, a measure for the

Training progress of the ANN, a measure of the utilization of the neurons of a layer or of another sub-area of the ANN, and / or a measure of the utilization of the neurons of the ANN as a whole.

One or more neurons are evaluated on the basis of a measure for their respective quantitative contributions to the determined quality. Measures with which the evaluated neurons are trained in the further course of the training, and / or positions of these neurons in the ANN are based on determined by the ratings of the neurons. These measures can then be carried out in further training. The established values of the

Neurons in the ANN can also continue to apply for the inference phase, i.e. for the later effective operation of the ANN after training.

In particular, for example, the measure for the quality can be evaluated as a weighted or unweighted sum of quantitative contributions from individual neurons.

It was recognized that in this way the wish can be taken into account during the training, the neurons of the ANN as well as the

Optimally utilize connections between these neurons. This wish can, for example, lead to an optimization goal for training the ANN. Should it become apparent during training that, despite the explicit request, certain neurons or connections between neurons are not optimally utilized, these neurons or connections can be deactivated during training. This has various advantages over a subsequent “pruning” after completing the training.

Should be early on during exercise

point out that certain neurons or connections between neurons are less relevant, these neurons or connections can be deactivated at an early stage. From this point in time, there is no further computing effort for these neurons or connections during training. It was recognized that a price has to be paid for each final deactivation or removal of a neuron or a connection in the form of the computational effort already invested in the training of this neuron or the connection. After the deactivation or removal, the knowledge from the training embodied in this neuron or in this connection is no longer used, ie the effort already invested is discarded. The amount of this effort is advantageously reduced, analogous to the practice in academic or professional training, to screen unsuitable candidates for the subject as early as possible. The desired end result, namely a fully trained and at the same time on the actually relevant neurons and Connections limited ANN, so can be obtained comparatively quickly.

The restriction to the actually relevant neurons and connections is in turn important for the efficient implementation of the fully trained ANN on hardware. Particularly when using KNNs in control units for vehicles, for example for driver assistance systems or systems for at least partially automated driving, the specification with regard to the available hardware is often already established before training of the ANN is started. The finished ANN is then limited in its size and complexity to these given hardware resources. At the same time, it must be used for inference, i.e. when evaluating input variables in

Real operation, get by with these hardware resources in order to deliver the required output variables in the response time specified for the respective application. Every deactivated neuron and every deactivated connection saves computational effort and therefore response time for every further inference.

The deactivation of neurons or connections basically represents an intervention in the ANN. By this intervention during the ongoing

Training process takes place, the training process can react to the intervention. Side effects of the intervention, such as overfitting on the training data, poor generalization of the ANN to unknown situations or an increased susceptibility to manipulation of the inference by presenting an "adverse example", can be significantly reduced. In contrast to, for example, the random deactivation of a certain percentage of neurons during training (“random dropout”), this does not mean that a proportion of the learned information corresponding to this percentage remains unused. The reason for this is that the deactivation of the neurons or connections is motivated from the outset by the lack of relevance of the neurons or connections in question for the quality.

Finally, the range of possible actions that can be taken in response to a small quantitative contribution of certain neurons to quality is expanded beyond the mere deactivation of these neurons. For example, further training can target such "weak" Neurons are focused so that they can still make a productive contribution to quality. This can be compared to the fact that in school education when problems arise in a certain subject, tutoring is usually used as the first measure, instead of immediately denying the pupil's talent for this subject.

The previous period advantageously comprises at least one epoch of the training, i.e. a period in which each of the available learning data sets, each comprising learning values for the input variables and associated learning values for the output variables, was used once. The determined quantitative contributions of the neurons to quality can then be better compared. It is quite possible that certain neurons of the ANN are “specialized” in a good treatment of certain situations occurring in the input variables, i.e. that these situations are particularly “suited” to these neurons. If a period is considered in which predominantly this

When situations occur, the performance of these neurons is rated higher than it really is, because in other situations the performance of other neurons can be significantly better. In the analogue of academic training, this corresponds to an examination for which the candidate is selective for one

has prepared a certain sub-area of the examination material and is "lucky" that this sub-area is selectively queried. The assessment then puts the other candidates at a disadvantage, whose knowledge is less detailed but covers the entire spectrum of the examination material much better. The

Consideration of at least one epoch corresponds to a fairer examination with a wide range of questions from the entire spectrum of the examination material.

In a particularly advantageous embodiment, a change in a cost function (loess function), the optimization of which the training of the ANN is aimed at, is included in the measure of quality over the past period. This ensures that the actions taken in response to the evaluation of the neurons do not conflict with the ultimate goal of training the ANN.

In a further particularly advantageous embodiment, quantitative contributions of neurons to the quality are weighted higher, the shorter they are The provision of these contributions is in the past. That way you can

It must be taken into account that knowledge learned from the ANN, like any other knowledge, can become out of date, so that new knowledge must take its place.

If, for example, starting from a current iteration step t, a time period is considered that goes back N iteration steps, then the quantitative contributions that a layer k of the ANN has made to the quality of the ANN in this period, for example an iteration progress, can be cumulated and / or aggregated become. In particular, it can also have a

Dropout tensor taken into account which neurons in the Send k were actually active in a given iteration step tn. For example, for a fully networked layer k of the ANN, the contributions of the respective active neurons to the quality can be determined over all previous N iteration steps in an "activation quality" summarize, which can be written, for example, as

M, contains the corresponding quality of all individual neurons in layer k and depends on the journal t.

Here L is the cost function (loess function). Accordingly, AL _{tn is} the difference in the cost function between the time step tn and the journal t. The effects of iteration steps that were different in time are normalized by the decay parameter g. The dropout tensor indicates for each layer k of the ANN and for each iteration step tn which neurons of the layer k were active or not in the iteration step tn. thus serves to take into account a possible temporary deactivation of individual neurons for individual iteration steps (dropout). It is not mandatory that a dropout be used when training the ANN. Depending on the type of layer k,] M ^ Be vectors or matrices.

The described "activation quality" can alternatively also be used as a function of the signals and weights express for all neurons of layer k. The signals are in the broadest sense activations of neurons, such as with weights weighted sums of inputs corresponding to the respective

Neurons are fed. can then be used, for example .write as

M = normf S / '^ Q ^) \

\ n = 0

Here x _t denotes the inputs that are fed to the KN N as a whole.

In each iteration step, the weights assigned to neurons in the ANN are changed by certain amounts in accordance with the cost function and the training strategy used (such as, for example, Stochastic Gradient Descent, SGD). In a further particularly advantageous embodiment, these amounts are reinforced with a multiplicative factor that is used for

Neurons with higher quantitative contributions to quality is lower than for neurons with lower quantitative contributions to quality. Neurons with a currently lower performance are therefore subjected to stronger learning steps with the aim of actively improving performance, analogous to

Tutoring. For example, the iteration steps of the weights

be performed. In a further particularly advantageous embodiment, neurons are temporarily deactivated during training with a probability which is higher for neurons with higher quantitative contributions to quality than for neurons with lower quantitative contributions to quality.

This measure also serves the targeted promotion of neurons whose quantitative contributions to quality are currently low. The temporary deactivation of the powerful neurons in particular forces the ANN to also involve the weaker neurons in the formation of the ultimate output variables. Accordingly, these weaker neurons also receive more feedback from the comparison of the output variables with the “ground truth” in the form of the learning values for the output variables. As a result, their performance tends to improve. The situation can be compared to teaching in a class with a heterogeneous performance level. If the teacher's questions only go to the strong students, the weaker students do not learn, so that the gap between the strong and the weaker students is cemented or even widened.

As a further consequence, this measure also makes the ANN more robust against failures of the powerful neurons, because precisely such situations are practiced by the ANN through the temporary deactivation.

For example, the dropout tensor for t> 0 over a Bernoulli

Function can be sampled. This in turn depends on the activation quality M, for a previous iteration step. This results in the probabilities p ^k that neurons of a layer k are activated

In a further particularly advantageous embodiment of the invention, neurons with higher quantitative contributions to quality are assigned higher values in the ANN than neurons with lower quantitative contributions to quality. The significance can manifest itself, for example, in the weight with which outputs from the relevant neurons are taken into account or whether the neurons are activated at all. This embodiment can be used in particular to compress the ANN to the part that is relevant for it. For this purpose, neurons whose quantitative contributions to quality meet a specified criterion can be deactivated in the ANN. The criterion can for example be formulated as an absolute criterion, such as a threshold value. However, the criterion can also be formulated, for example, as a relative criterion, such as a deviation of the quantitative contributions to quality from the quantitative contributions of other neurons or from a summary statistic thereof. A

Summary statistics can include, for example, a mean, median, and / or standard deviations.

Unlike neurons, whose outputs are only underweighted, deactivated neurons can be completely saved when implementing the fully trained ANN on hardware. The more the ANN is compressed, the fewer hardware resources are required for the final implementation.

The same applies if unimportant connections between neurons in the ANN are deactivated. Each of these connections costs additional computing time for inference during operation of the ANN, because the output of the neuron at one end of the connection must be multiplied by the weight of the connection and is then included in the activation of the neuron at the other end of the connection. If the weight of the connection is different from zero, these arithmetic operations occur and always take the same time, no matter how close it is

Weight may be zero and how little this is taken into account

Connection ultimately results in the output of the ANN. Therefore, in a further particularly advantageous embodiment, connections between

Neurons whose weights meet a given criterion are deactivated in the ANN. Analogously to the criterion for the quantitative contributions of neurons, the criterion can be formulated absolutely or relatively.

In a further particularly advantageous embodiment, the number of neurons activated in the ANN and / or in a sub-area of the ANN is reduced from a first number to a predetermined second number by adding neurons deactivated with the lowest quantitative contributions. For example, the hardware used can dictate the maximum complexity of the ANN.

The invention also relates to a method for implementing an ANN on a predetermined arithmetic unit. In this method, a model of the ANN is trained in a training environment outside the arithmetic unit using the method described above. When the training is completed, activated neurons and connections between neurons are implemented on the arithmetic unit.

The specified arithmetic unit can be designed, for example, to be built into a control unit for a vehicle, and / or it can be designed to be supplied with energy from the on-board network of a vehicle. The available space and the heat budget in the control unit, or the amount of electricity available, then limit the hardware resources available for inferring the ANN.

In contrast, the training environment can be equipped with significantly more resources. For example, a physical or virtual computer with a powerful graphics processor (GPU) can be used. Little to no preliminary thought is required to begin exercising; the model should only have a certain minimum size with which the problem to be solved can probably be represented with sufficient accuracy.

The method described above makes it possible to determine within the training environment which neurons and connections between neurons are important. Based on this, the ANN can be compressed for implementation on the arithmetic unit. As described above, this can also be done automatically within the training environment.

It can therefore generally be advantageous on the basis of a predetermined one

Arithmetic logic unit, which hardware resources for a given number of neurons, layers of neurons and / or connections between neurons , a model of the ANN can be selected whose number of neurons, layers of neurons and / or connections between neurons exceeds the predetermined number. The compression ensures that the trained ANN ultimately fits the specified hardware. The aim here is to ensure that those neurons and connections between neurons that are ultimately implemented on the hardware are also those that are most important for the inference in the operation of the ANN.

The invention also relates to another method. With this one

In the method, an artificial neural network, ANN, is initially trained with the method described above for training, and / or implemented on an arithmetic unit using the method described above for implementation. The ANN is then operated by supplying it with one or more input variables. Depending on the output variables supplied by the ANN, a vehicle, a robot, becomes a

Quality control system and / or a system for monitoring an area on the basis of sensor data.

In the method described, in particular an ANN can be selected which is designed as a classifier and / or regressor for physical measurement data recorded with at least one sensor. In applications in which only limited hardware, a limited amount of energy or a limited installation space is available, it then enables a meaningful evaluation of the physical measurement data that can be generalized to many situations using artificial intelligence. The sensor can be, for example, an imaging sensor, a radar sensor, a lidar sensor or an ultrasonic sensor.

In particular, the methods can be implemented entirely or partially in software that brings about the immediate customer benefit that an ANN delivers better results in relation to the hardware expenditure and the energy consumption for the inference in its operation. The invention therefore also relates to a computer program with machine-readable instructions which, when they are on a computer, and / or on a control device, and / or on a Embedded system are executed, cause the computer, the control unit, and / or the embedded system to execute one of the methods described. Control devices and embedded systems can therefore be regarded as computers at least in the sense that their behavior is characterized in whole or in part by a computer program. The term “computer” thus encompasses any device for processing that can be specified

Calculation rules. These calculation rules can be in the form of software, or in the form of hardware, or also in a mixed form of software and hardware.

The invention also relates to a machine-readable one

Data carrier or a download product with the computer program. A download product is a product that can be transmitted over a data network, i.e. digital product downloadable by a user of the data network that

for example, it can be offered for immediate download in an online shop.

Furthermore, a computer with the computer program with which

be equipped with a machine-readable data carrier or with the download product, and / or specifically set up in any other way to carry out one of the described methods. Such a specific device can be implemented, for example, with field-programmable gate arrangements (FPGAs) and / or application-specific integrated circuits (ASICs).

Further measures improving the invention are shown in more detail below together with the description of the preferred exemplary embodiments of the invention with reference to figures.

Embodiments

It shows:

FIG. 1 exemplary embodiment of the method 100 for training an ANN 1; FIG. 2 exemplary embodiment of the method 200 for implementing the KNN 1 on an arithmetic unit 4;

FIG. 3 Exemplary compression of an ANN 1.

FIG. 1 shows an exemplary embodiment of the method 100. The method 100 is embedded in the training of the ANN 1, which takes place in a known manner, which includes the loading of the ANN 1 with learning values for input variables, the comparison of the output variables formed by the ANN 1 with the Includes learning values for the output variables as well as the change in parameters within the ANN in accordance with the cost function. These steps are the

Not shown in FIG. 1 for the sake of clarity.

In step 110, at any point in time and in any phase of the training process, a measure for the quality 11 that the ANN achieved within a predetermined previous period is determined. According to block 111, in particular the change in the cost function used in training the ANN 1 can be included in the measure for the quality 11.

In step 120, several neurons 2 are evaluated using a measure for their respective quantitative contributions 21 to the previously determined quality 11. In particular, according to block 121, these contributions 21 can be weighted higher, the shorter the time since the contributions 21 were made. These contributions 21 can for example from the previously described in detail

Activation quality] M ^ can be determined. In particular, values of the activation quality] M ^ can be used directly as contributions 21.

On the basis of the evaluations 120a created in this way, measures 22 for the further training of the evaluated neurons 2 and / or position values 23 of these evaluated neurons 2 in the ANN 1 are determined in step 130.

Specifically, amounts by which the weights of neurons 2 are changed in at least one training step can, according to block 131, with a multiplicative factor are strengthened, which for stronger quality 11

contributing neurons 2 is lower than for weaker contributing neurons 2 to quality 11.

According to block 132, neurons 2 can be temporarily deactivated during training, the probability of such a deactivation being higher for neurons 2 with higher quantitative contributions 21 to quality 11 than for neurons 2 with lower quantitative contributions 21 to quality 11.

According to block 133, neurons 2 with higher quantitative contributions 21 to quality 11 can be assigned higher values in the KN N 1 than neurons 2 with lower quantitative contributions 21.

This can include, for example, that neurons 2, the quantitative contributions 21 of which meet a predetermined criterion, are deactivated according to sub-block 133a. According to sub-block 133b, connections 25 between neurons 2, the weights of which meet a predefined criterion, can also be deactivated. According to sub-block 133c, the number of activated neutrons can be reduced to a predetermined number in a targeted manner by

Neurons 2 with the lowest quantitative contributions 21 to quality 11 are deactivated.

FIG. 2 shows an exemplary embodiment of the method 200 for implementing the ANN 1 on a predetermined arithmetic unit 4. In the optional step 205a, an arithmetic unit 4 with limited resources with regard to the number of neurons 2, layers 3a, 3b with neurons 2, and / or connections 25 between neurons 2, are preselected. The model la of the ANN can then be selected so that it has a number of neurons 2, layers 3a, 3b or connections 25 that are above the respective limit of the

Arithmetic unit 4 is located. In particular, according to optional step 206, an arithmetic unit 4 can be selected which is designed to be installed in a control device for a vehicle and / or to be supplied with energy from the on-board network of a vehicle. In step 210, the ANN 1 is trained according to the method 100 described above. Upon completion of training 210 activated neurons 2 and

Connections 25 between neurons 2 are in step 220 on the

Arithmetic unit 4 implemented. As previously described, during exercise 210 compression may be too large an ANN for limited

Make hardware resources "suitable". If this is not sufficient, the neurons 2, layers 3a, 3b, or connections 25 can be read out further after the training 210 has been completed.

FIG. 3 shows the effect of compression on an exemplary ANN 1. This ANN 1 comprises two layers 3a and 3b each with four neurons 2, the exemplary quantitative contributions 21 of which to the quality 11 of the ANN 1 are indicated overall.

In the state shown in Figure 3a, the two layers 3a and 3b are still fully networked with one another, i.e. each neuron 2 of the first layer 3a is connected to each neuron 2 of the second layer 3b, and all neurons 2 are still active.

In the state shown in FIG. 3b, the rule was applied that all neurons 2 with quantitative contributions 21 of less than 0.5, as well as the connections 25 leading to these neurons 2, are deactivated. The neurons outlined in dashed lines in the first layer 3a have fallen victim to this rule. As a result, the number of connections 25 that are still active has been halved from 16 to 8, which correspondingly reduces the computational effort for the inference during operation of the ANN 1.

Claims

Expectations

1. Method (100) for training an artificial neural

Network, ANN (1), which comprises a large number of neurons (2), with the steps:

• It becomes a measure of the quality (11), the ANN (1), and / or a

Sub-area of the ANN (1), achieved overall within a previous period, determined (110);

• one or more neurons (2) are evaluated (120) using a measure for their respective quantitative contributions (21) to the determined quality (11);

Measures (22) with which the evaluated neurons (2) are trained in the further course of the training, and / or place values (23) of these neurons (2) in the ANN (1), are based on the evaluations (120a) of the neurons (2) established (130).

2. The method (100) according to claim 1, wherein the measure of the quality is a measure of the training progress of the ANN (1), a measure of the utilization of the neurons of a layer or another sub-area of the ANN (1), and / or a Measure of the total utilization of the neurons of the ANN (1).

3. The method (100) according to any one of claims 1 to 2, wherein the measure for the quality (11) is evaluated as a weighted or unweighted sum of quantitative contributions (21) of individual neurons (2).

4. The method (100) according to any one of claims 1 to 3, wherein a

Change of a cost function, the optimization of which the training of the ANN (1) is aimed at, over the past period is included in the measure for the quality (11) (111).

5. The method (100) according to any one of claims 1 to 4, wherein quantitative contributions (21) of neurons (2) to the quality (11) are weighted higher (121), the shorter the time since these contributions (21) were made .

6. The method (100) according to any one of claims 1 to 5, wherein amounts by which the weights assigned to neurons (2) in the ANN (1) are changed in at least one training step are amplified with a multiplicative factor ( 131), which is lower for neurons (2) with higher quantitative contributions (21) than for neurons (2) with lower quantitative contributions (21).

7. The method (100) according to any one of claims 1 to 6, wherein neurons (2) are temporarily deactivated (132) with a during training

Probability that is higher for neurons (2) with higher quantitative contributions (21) than for neurons (2) with lower quantitative contributions (21).

8. The method (100) according to any one of claims 1 to 7, wherein the previous period comprises at least one epoch of training.

9. The method (100) according to any one of claims 1 to 8, wherein neurons (2) with higher quantitative contributions (21) are assigned higher values in the ANN (1) than neurons (2) with lower quantitative ones

Contributions (21).

10. The method (100) according to claim 9, wherein neurons (2) whose quantitative contributions (21) meet a predetermined criterion are deactivated (133a) in the ANN (1).

11. The method (100) according to any one of claims 9 to 10, wherein

Connections (25) between neurons (2), the weights of which meet a predetermined criterion, are deactivated in the ANN (1) (133b).

12. The method (100) according to any one of claims 9 to 11, wherein the number of neurons (2) activated in the ANN (1) and / or in a sub-area of the ANN (1) from a first number to a predetermined second number is reduced by deactivating neurons (2) with the lowest quantitative contributions (21) (133c).

13. Method (200) for implementing a KN N (1) on a given arithmetic logic unit (4) with the steps:

• a model (la) of the ANN (1) is used in a training environment (5)

trained (210) outside of the arithmetic logic unit (4) with the method (100) according to one of claims 1 to 12;

• at the end of the training (210) activated neurons (2) and

Connections (25) between neurons (2) are implemented (220) on the arithmetic unit (4).

14. The method (200) according to claim 13, wherein an arithmetic unit (4) is selected (205a), which hardware resources for a predetermined number of neurons (2), layers (3a, 3b) of neurons (2) and / or connections ( 25) between neurons (2), and where a model (la) of the ANN (1) is selected (205b), its number of neurons (2), layers (3a, 3b) of neurons (2) and / or connections (25) between neurons (2) exceeds the specified number.

15. Procedure with the steps:

• an artificial neural network, KNN (1), is trained with the method (100) according to one of claims 1 to 12, and / or implemented with the method (200) according to one of claims 13 to 14;

• the ANN (1) is operated by giving it one or more

Input variables are fed;

• Depending on the output variables supplied by the ANN (1), a vehicle, a robot, a quality control system and / or a system for monitoring an area on the basis of sensor data is controlled.

16. Computer program, containing machine-readable instructions which, when executed on a computer and / or on a control device and / or on an embedded system, the computer, the control device, and / or cause the embedded system to carry out a method according to one of claims 1 to 15.

17. Machine-readable data carrier and / or download product with the computer program according to claim 16.

18. Computer equipped with the computer program according to claim 16, with the machine-readable data carrier or with the download product according to claim 17, and / or otherwise specifically set up to carry out a method (100, 200) according to one of claims 1 to 15 .