Information processing method and apparatus, and device and storage medium

Abstract

The embodiments of the present disclosure relate to an information processing method and apparatus, and a device and a storage medium. The method provided herein comprises: acquiring a training input concerning a strategy model; on the basis of a performance indicator of the strategy model during training, determining a maximum output length of the strategy model; acquiring a first result which is output by the strategy model on the basis of the maximum output length; and on the basis of a comparison between the first result and a second result, adjusting parameters of the strategy model, wherein the second result is generated by a first reference model on the basis of the training input. In this way, the embodiments of the present disclosure can use a continuous maximization strategy to dynamically adjust an output length, and generate more detailed and analytical content, thereby improving the performance of a model.

Description

Information processing method, device, equipment and storage medium Technical Field

[0001] Example embodiments of the present disclosure generally relate to the field of computers, and more particularly, to an information processing method, apparatus, device, and computer-readable storage medium. Background Art

[0002] With the development of computer technology, natural language processing has become an important task. Traditional language models rely on the supervised fine-tuning (SFT) training process. This process tends to produce models that perform well on specific tasks, but these models lack in the generalization and depth of responses. Although SFT reduces hallucinations, the generated responses are often short and lack substantive analysis, which hinders subsequent reinforcement learning training from human feedback. In addition, these models are often trained on databases that fail to fully utilize their linguistic and cognitive capabilities, resulting in poor performance in tasks that require high-level reasoning and detailed analysis.

[0003] Summary of the Invention

[0004] In a first aspect of the present disclosure, a method for information processing is provided. The method comprises: obtaining training input for a policy model; determining a maximum output length of the policy model based on a performance indicator of the policy model during training; obtaining a first result output by the policy model based on the maximum output length; and adjusting parameters of the policy model based on a comparison of the first result with a second result generated by a first reference model based on the training input.

[0005] In a second aspect of the present disclosure, a device for information processing is provided. The device includes: an input acquisition module configured to acquire training input for a policy model; a length determination module configured to determine a maximum output length of the policy model based on a performance indicator of the policy model during training; a result acquisition module configured to acquire a first result output by the policy model based on the maximum output length; and a parameter adjustment module configured to adjust parameters of the policy model based on a comparison of the first result and a second result, the second result being generated by a first reference model based on the training input.

[0006] In a third aspect of the present disclosure, an electronic device is provided. The device includes at least one processing unit; and at least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit. When executed by the at least one processing unit, the instructions cause the device to perform the method of the first aspect.

[0007] In a fourth aspect of the present disclosure, a computer-readable storage medium is provided, wherein a computer program is stored on the computer-readable storage medium, and the computer program can be executed by a processor to implement the method of the first aspect.

[0008] In a fifth aspect of the present disclosure, a computer program product is provided, which includes computer-executable instructions, which, when executed by a processor, implement the method according to the first aspect of the present disclosure.

[0009] It should be understood that the content described in this summary section is not intended to limit the key features or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0011] FIG1 illustrates the architecture of an example training system capable of implementing some embodiments of the present disclosure;

[0012] FIG2 shows a flowchart of an information processing process according to some embodiments of the present disclosure;

[0013] FIG3 shows a schematic structural block diagram of an example apparatus for information processing according to some embodiments of the present disclosure; and

[0014] FIG4 illustrates a block diagram of an electronic device capable of implementing various embodiments of the present disclosure. DETAILED DESCRIPTION

[0015] The following describes embodiments of the present disclosure in more detail with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of protection of the present disclosure.

[0016] It should be noted that the titles of any section / subsection provided herein are not limiting. Various embodiments are described throughout this document, and any type of embodiment may be included under any section / subsection. Furthermore, the embodiments described in any section / subsection may be combined in any manner with any other embodiments described in the same section / subsection and / or in different sections / subsections.

[0017] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". Other explicit and implicit definitions may be included below. The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may be included below.

[0018] The embodiments of the present disclosure may involve user data, data acquisition and / or use, etc. These aspects shall comply with the corresponding laws, regulations and relevant provisions. In the embodiments of the present disclosure, all data collection, acquisition, processing, processing, forwarding, use, etc. are carried out on the premise that the user is aware of and confirms them. Accordingly, when implementing the various embodiments of the present disclosure, the types, scope of use, and usage scenarios of the data or information that may be involved should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with the relevant laws and regulations. The specific notification and / or authorization method may vary according to the actual situation and application scenario, and the scope of the present disclosure is not limited in this respect.

[0019] Any processing of personal information in this specification and the embodiments will be conducted only with a legitimate basis (e.g., with the consent of the subject of personal information or as necessary for the performance of a contract) and only within the prescribed or agreed scope. A user's refusal to process personal information other than that required for basic functions will not affect the user's use of basic functions.

[0020] As discussed above, the responses generated by SFT-based models are often short and lack substantive analysis, which hinders subsequent reinforcement learning from human feedback. Furthermore, these models are often trained on databases that fail to fully exploit their linguistic and cognitive capabilities, resulting in poor performance on tasks that require high-level reasoning and detailed analysis.

[0021] The embodiments of the present disclosure propose a scheme for information processing. According to the scheme, a training input for a policy model can be obtained. Furthermore, the maximum output length of the policy model can be determined based on the performance indicators of the policy model during the training process. Accordingly, a first result output by the policy model according to the maximum output length can be obtained. Furthermore, the parameters of the policy model can be adjusted based on a comparison of the first result and a second result, where the second result is generated by a first reference model based on the training input. In this way, the embodiments of the present disclosure can utilize a continuous maximization strategy to dynamically adjust the output length and generate more detailed and analytical content, thereby improving model performance.

[0022] Various example implementations of this solution are described in detail below in conjunction with the accompanying drawings.

[0023] Example training architecture

[0024] Figure 1 illustrates the architecture of an example training system 100 according to some embodiments of the present disclosure. As shown in Figure 1 , the training system 100 may include multiple models, namely, a reference model 110, a policy model 115, a reward model 120, and an evaluation model 125. The policy model 115 is the target model to be trained.

[0025] In some examples, the reference model 110 can serve as a pre-trained benchmark, providing a baseline for behavior comparison and ensuring that the output of the policy model remains consistent with the expected behavior during training. During training, the parameters of the reference model 110 can remain unchanged.

[0026] In some embodiments, the policy model 115 (also called the actor model) may be responsible for generating actions or decisions, selecting the best action based on the current state, and self-optimizing based on feedback from the reward model and the evaluation model to improve decision quality.

[0027] In some embodiments, reward model 120 may be used to evaluate and quantify the quality of responses generated by the policy model, providing performance feedback to policy model 115 .

[0028] In some embodiments, the evaluation model (also called the critic model) 125 can be used to estimate the state value under the current policy, providing guidance to the policy model 115 through the value function to help it understand the expected rewards of taking different actions in a specific state.

[0029] Specifically, as shown in FIG1 , the guide word sequence 105 can be provided to the above four models, and corresponding output information can be obtained from each of them. Specifically, the reference model 110 can provide a reference model output 130 ; the policy output model 135 can output a policy model output 135 based on the continuous maximization mechanism described below; the reward model 120 can provide a reward score 140 , and the evaluation model 125 can output a criticism score 145 .

[0030] Furthermore, the training system 100 may determine difference information 150 (e.g., KL divergence) between the reference model output 130 and the policy model output 135, and may further determine a reward function 155 based on the difference information 150 and the reward score 140. Accordingly, the training system 100 may determine an advantage function 160 for adjusting the model parameters of the policy model 115 based on the reward function 155 and the criticism value 145, thereby completing the training process of the policy model 115.

[0031] The specific training process of the strategy model 115 will be described in detail below.

[0032] Example Process

[0033] FIG2 shows a flowchart of an example process 200 for information processing according to some embodiments of the present disclosure. The process 200 shown in FIG2 will be described below with reference to FIG1 .

[0034] As shown in FIG2 , at block 210 , the training system 100 obtains training input for the policy model. Taking FIG1 as an example, the training system 100 may obtain prompt word training as training input for the policy model 115 .

[0035] In block 220 , the training system 100 determines a maximum output length of the policy model based on the performance metrics of the policy model during training.

[0036] In some embodiments, unlike a traditional fixed number of iterations, the training system 100 may determine the maximum output length of the policy model 115 based on a continuous maximization strategy.

[0037] In some embodiments, training system 100 may determine a performance metric for policy model 115 at time step t. In some embodiments, such a performance metric may include, for example, reward score 140 and / or critic value 145 (also referred to as critic value function loss) as shown in FIG1 .

[0038] In some embodiments, the training system 100 may determine whether the performance indicator satisfies a preset condition. Such a preset condition may, for example, be related to a change in the performance indicator within a predetermined number of time steps.

[0039] For example, the training system 100 may determine a change in the performance indicator over a plurality of time steps. Further, the training system 100 may determine whether the magnitude of the change is greater than a threshold, or the training system 100 may determine that the change indicates a degradation in processing performance.

[0040] If the magnitude of the change is greater than a threshold (e.g., the performance indicator fluctuates significantly), the training system 100 may determine that the performance indicator does not meet the preset condition. As another example, if the performance indicator indicates that the processing performance of the model has degraded, the training system 100 may determine that the performance indicator does not meet the preset condition.

[0041] When it is determined that the performance indicator does not meet the preset condition, the training system 100 may keep the maximum output length unchanged, that is, determine the output length of the previous time step as the maximum output length of the current time step.

[0042] On the contrary, if the performance indicator meets the preset conditions, the training system 100 can use the prediction module to process the performance indicator to determine the maximum output length. This process can be expressed as follows:

[0043] l (t) =f CM (R t (s t , a t ), V φ (s t )) (1)

[0044] Among them, the new maximum output length l (t) is based on the reward model score R t (s t , a t ) and the critical value function V φ (s t ), and through the logistic regression function.f CM Calculated. φ is the parameter of the critical value function, S t represents the current state at time step t, a t represents the action taken at time step t.

[0045] In some embodiments, the logistic regression function may be implemented based on, for example, a multi-layer perceptron (MLP) to achieve prediction of the maximum output length.

[0046] As an example, based on the continuous maximization control process above, if the reward score is stable and the critic value function has little fluctuation, the output length of the model will gradually increase. Conversely, if the performance indicators show instability or decline (for example, the critic loss oscillates significantly), the output length may remain unchanged.

[0047] Thus, the embodiments of the present disclosure can adaptively adjust the maximum output length of the model based on performance parameters, thereby expanding the analysis and reasoning capabilities of the model and gradually and effectively improving its performance based on real-time feedback.

[0048] In block 230 , the training system 100 obtains a first result output by the policy model according to the maximum output length.

[0049] In some embodiments, to prevent policy model 115 from outputting an output that is not in accordance with the maximum output length, training system 100 may prevent policy model 115 from outputting an end identifier, such as [EOS] or [EOS], to ensure that policy model 115 can process the request in accordance with the maximum output length. Specifically, training system 100 may, for example, use system prompts to cause policy model 115 to fully process the request without outputting any end identifier.

[0050] At block 240 , the training system 100 adjusts parameters of the policy model based on a comparison of the first result and a second result generated by the first reference model based on the training input.

[0051] 1 , training system 100 may determine difference information 150 based on a comparison of a first result (e.g., policy model output 135) and a second result (e.g., reference model output 130), and further determine advantage function 160. Furthermore, training system 100 may adjust parameters of policy model 115 based on advantage function 160.

[0052] Exemplarily, the process can be expressed as:

[0053] Among them, formula (2) means that at time step t, given the current state s t and the length l of the previous output (t-1) , according to the strategy Take action a t Here, π is the policy function, θ coarse is a parameter of the policy model 115. Formula (3) represents the update amount Δ θcoarse is the logarithmic gradient of the current policy and the advantage function A π (s t , a t ). α is the learning rate, represents the gradient operation, and the advantage function A π (s t , a t ) is used to estimate the action a taken under the current strategy t The average advantage of taking this action compared to other possible actions.

[0054] In this way, by employing a CM (Continuous Maximization) approach in the policy model, embodiments of the present disclosure create a detailed and extensive knowledge base that significantly enhances the analytical and reasoning output of the model.

[0055] In some embodiments, in order to reduce the redundancy of the policy model output content, the embodiments of the present disclosure may also perform a second stage of model fine-tuning.

[0056] Specifically, the training system 100 may obtain a first set of parameters of the trained policy model. For example, the first set of parameters may be determined based on the training process described above.

[0057] Furthermore, the training system 100 may determine a set of target parameters for the policy model 115 by updating the first set of parameters using a second set of parameters of a second reference model.

[0058] In some embodiments, second reference model 115 may be an SFT model, which may have the same architecture as policy model 115. In this case, training system 100 may combine model parameters of policy model 115 and model parameters of the SFT model to determine final model parameters.

[0059] In some embodiments, the training system 100 may determine the set of target parameters based on a weighted sum of the first set of parameters and the second set of parameters. For example, the process may be represented as follows:

[0060] θ fine =γθ coarse +(1-γ)θ SFT (4)

[0061] θ coarse represents the model parameters of the strategy model after the first stage of training; θ SFT represents the model parameters of the SFT model; γ represents the weight coefficient.

[0062] Thus, embodiments of the present disclosure can further utilize the extensive knowledge base obtained from the first stage of training through parameter fusion while maintaining the accuracy and context-specific capabilities of the SFT model, thereby enhancing the overall analysis and reasoning capabilities in conversational interactions.

[0063] Example devices and equipment

[0064] Embodiments of the present disclosure also provide corresponding apparatuses for implementing the above-described methods or processes. FIG3 shows a schematic block diagram of an example apparatus 300 for information processing according to certain embodiments of the present disclosure. Apparatus 300 may be implemented as or included in an electronic device. Each module / component in apparatus 300 may be implemented by hardware, software, firmware, or any combination thereof.

[0065] As shown in Figure 3, the device 300 includes an input acquisition module 310, which is configured to obtain training input for the strategy model; a length determination module 320, which is configured to determine the maximum output length of the strategy model based on the performance indicators of the strategy model during the training process; a result acquisition module 330, which is configured to obtain a first result output by the strategy model according to the maximum output length; and a parameter adjustment module 340, which is configured to adjust the parameters of the strategy model based on a comparison of the first result and a second result, where the second result is generated by a first reference model based on the training input.

[0066] In some embodiments, the length determination module 320 is further configured to: determine whether the performance indicator meets a preset condition at the target time step of the training process; and in response to the performance indicator meeting the preset condition, use a prediction module to process the performance indicator to determine the maximum output length.

[0067] In some embodiments, the length determination module 320 is further configured to: in response to the performance indicator not meeting the preset condition, determine the maximum output length as the output length of the previous time step.

[0068] In some embodiments, the length determination module 320 is further configured to: determine a change in the performance indicator within multiple time steps; and in response to the magnitude of the change being greater than a threshold or the change indicating a decrease in processing performance, determine that the performance indicator does not meet the preset condition.

[0069] In some embodiments, the parameter adjustment module 340 is further configured to: determine an advantage function based on a comparison of the first result and the second result; and adjust the parameters of the policy model based on the advantage function.

[0070] In some embodiments, the device 300 also includes a fusion module configured to: obtain a first set of parameters of the trained policy model; and determine a set of target parameters of the policy model by updating the first set of parameters using a second set of parameters of a second reference model.

[0071] In some embodiments, the fusion module is further configured to determine the set of target parameters based on a weighted sum of the first set of parameters and the second set of parameters.

[0072] In some embodiments, the performance metric comprises a reward model score and / or a critic value function loss.

[0073] FIG4 shows a block diagram of an electronic device 400 in which one or more embodiments of the present disclosure may be implemented. It should be understood that the electronic device 400 shown in FIG4 is merely exemplary and should not be construed as limiting the functionality and scope of the embodiments described herein. The electronic device 400 shown in FIG4 can be used to implement the training system 100.

[0074] As shown in FIG4 , electronic device 400 is a general-purpose electronic device. Components of electronic device 400 may include, but are not limited to, one or more processors or processing units 410, memory 420, storage device 430, one or more communication units 440, one or more input devices 450, and one or more output devices 460. Processing unit 410 may be a real or virtual processor and is capable of performing various processes according to programs stored in memory 420. In a multi-processor system, multiple processing units execute computer-executable instructions in parallel to enhance the parallel processing capabilities of electronic device 400.

[0075] The electronic device 400 typically includes a plurality of computer storage media. Such media can be any accessible media that can be obtained by the electronic device 400, including but not limited to volatile and non-volatile media, removable and non-removable media. The memory 420 can be a volatile memory (e.g., registers, cache, random access memory (RAM)), a non-volatile memory (e.g., read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. The storage device 430 can be a removable or non-removable medium and can include a machine-readable medium, such as a flash drive, a disk, or any other medium that can be used to store information and / or data and can be accessed within the electronic device 400.

[0076] The electronic device 400 may further include additional removable / non-removable, volatile / non-volatile storage media. Although not shown in FIG. 4 , a disk drive for reading from or writing to a removable, non-volatile disk (e.g., a “floppy disk”) and an optical drive for reading from or writing to a removable, non-volatile optical disk may be provided. In these cases, each drive may be connected to a bus (not shown) by one or more data media interfaces. The memory 420 may include a computer program product 425 having one or more program modules configured to perform various methods or actions of various embodiments of the present disclosure.

[0077] The communication unit 440 enables communication with other electronic devices via a communication medium. Additionally, the functions of the components of the electronic device 400 can be implemented in a single computing cluster or multiple computing machines that can communicate via a communication connection. Thus, the electronic device 400 can operate in a networked environment using a logical connection with one or more other servers, a network personal computer (PC), or another network node.

[0078] Input device 450 may be one or more input devices, such as a mouse, keyboard, or trackball. Output device 460 may be one or more output devices, such as a display, a speaker, or a printer. Electronic device 400 may also communicate with one or more external devices (not shown) via communication unit 440 as needed, such as a storage device, a display device, or the like, with one or more devices that allow a user to interact with electronic device 400, or with any device that allows electronic device 400 to communicate with one or more other electronic devices (e.g., a network card, a modem, etc.). Such communication may be performed via an input / output (I / O) interface (not shown).

[0079] According to an exemplary implementation of the present disclosure, a computer-readable storage medium is provided, on which computer-executable instructions are stored, wherein the computer-executable instructions are executed by a processor to implement the method described above. According to an exemplary implementation of the present disclosure, a computer program product is also provided, which is tangibly stored on a non-transitory computer-readable medium and includes computer-executable instructions, and the computer-executable instructions are executed by a processor to implement the method described above.

[0080] Various aspects of the present disclosure are described herein with reference to flowcharts and / or block diagrams of methods, apparatuses, devices, and computer program products implemented according to the present disclosure. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer-readable program instructions.

[0081] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, such that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device is generated that implements the functions / actions specified in one or more blocks in the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, where these instructions cause the computer, programmable data processing device, and / or other device to operate in a specific manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing various aspects of the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0082] Computer-readable program instructions can be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0083] The flow charts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple implementations of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a part for a module, program segment or instruction, and a part for a module, program segment or instruction comprises one or more executable instructions for realizing the logical function of the specification. In some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be realized by a special hardware-based system that performs the function or action of the specification, or can be realized by a combination of special hardware and computer instructions.

[0084] While various implementations of the present disclosure have been described above, the foregoing description is intended to be illustrative, not exhaustive, and not limited to the disclosed implementations. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described implementations. The terminology used herein is selected to best explain the principles of the implementations, their practical applications, or improvements to existing technologies, or to enable others skilled in the art to understand the various implementations disclosed herein.

Claims

1. An information processing method, comprising: Get training input for the policy model; Determining a maximum output length of the strategy model based on a performance indicator of the strategy model during training; Obtaining a first result output by the strategy model according to the maximum output length; as well as Parameters of the policy model are adjusted based on a comparison of the first result and a second result generated by a first reference model based on the training input.

2. The method of claim 1, wherein determining the maximum output length of the policy model comprises: At a target time step of the training process, determining whether the performance indicator meets a preset condition; as well as In response to the performance indicator satisfying the preset condition, the performance indicator is processed using a prediction module to determine the maximum output length.

3. The method of claim 2, wherein determining the maximum output length of the policy model comprises: In response to the performance indicator not satisfying the preset condition, the maximum output length is determined as the output length of the previous time step.

4. The method according to claim 2, wherein determining whether the performance indicator meets a preset condition comprises: determining a change in the performance indicator over a plurality of time steps; as well as In response to the magnitude of the change being greater than a threshold or the change indicating a decrease in processing performance, it is determined that the performance indicator does not satisfy the preset condition.

5. The method of claim 1 , wherein adjusting the parameters of the policy model based on the comparison of the first result and the second result comprises: determining an advantage function based on a comparison of the first result and the second result; as well as Based on the advantage function, the parameters of the policy model are adjusted.

6. The method according to claim 1, further comprising: Obtaining a first set of parameters of the trained policy model; as well as A set of target parameters for the policy model is determined by updating the first set of parameters using a second set of parameters of a second reference model.

7. The method of claim 6, wherein determining a set of target parameters of the policy model by updating the first set of parameters using a second set of parameters of a second reference model comprises: The set of target parameters is determined based on a weighted sum of the first set of parameters and the second set of parameters.

8. The method of claim 1, wherein the performance metric comprises a reward model score and / or a critic value function loss.

9. An apparatus for information processing, comprising: An input acquisition module configured to acquire training input for a policy model; a length determination module, configured to determine a maximum output length of the strategy model based on a performance indicator of the strategy model during training; A result acquisition module is configured to acquire a first result output by the strategy model according to the maximum output length; as well as A parameter adjustment module is configured to adjust the parameters of the policy model based on a comparison between the first result and a second result, where the second result is generated by the first reference model based on the training input.

10. An electronic device, comprising: at least one processing unit; as well as At least one memory, the at least one memory being coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions causing the electronic device to perform the method according to any one of claims 1 to 8 when executed by the at least one processing unit.

11. A computer-readable storage medium having a computer program stored thereon, wherein the computer program can be executed by a processor to implement the method according to any one of claims 1 to 8.

12. A computer program product comprising computer executable instructions, wherein the computer executable instructions, when executed by a processor, implement the method according to any one of claims 1 to 8.

Images