Learning system

Abstract

To qualify the output of a machine learning model.SOLUTION: A learning system comprises: learning means for learning a machine learning model to personalize a user using data acquired from the user; and limiting means for limiting the output of the machine learning model based on a comparison result between the machine learning model and a predetermined standard model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] TECHNICAL FIELD This disclosure relates to the technical field of learning systems. [Background technology]

[0002] It is known that models based on machine learning are used in devices that assist driving of automobiles, etc. For example, Patent Document 1 discloses a technology that uses the characteristics of a driver during manual driving to train a driving assistance device used during automated driving. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2019 / 058460 Summary of the Invention [Problem to be solved by the invention]

[0004] When training a machine learning model based on individual characteristics, there is a risk that inappropriate training may occur unintentionally.

[0005] This disclosure has been made in consideration of the above-mentioned problems, and aims to provide a learning system that can make the output of a machine learning model appropriate. [Means for solving the problem]

[0006] A learning system according to one aspect of the present disclosure includes a learning means for learning a machine learning model using data acquired from a user to personalize the model to the user, and a restriction means for restricting the output of the machine learning model based on a comparison result between the machine learning model and a predetermined standard model. [Brief explanation of the drawings]

[0007] [Figure 1]1 is a block diagram showing the configuration of a learning system according to a first embodiment. [Figure 2] 3 is a flowchart showing the flow of operations performed by the learning system according to the first embodiment. [Figure 3] 1 is a radar chart showing an example of output restriction of a machine learning model by the learning system according to the first embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0008] Hereinafter, an embodiment of a learning system will be described with reference to the drawings.

[0009] First Embodiment The first learning system will be described with reference to FIGS.

[0010] (System Configuration) First, the configuration of the learning system according to the first embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the configuration of the learning system according to the first embodiment.

[0011] In FIG. 1, a learning system 1 according to the first embodiment is configured to include a machine learning model 10 and a processing device 20.

[0012] The machine learning model 10 is a model that is trained to be personalized for a user using data acquired from the user (i.e., trained specifically for the individual user). The machine learning model 10 may be, for example, a model consisting of a neural network trained by deep learning. The machine learning model 10 may be, for example, a model used for driving assistance for automobiles and the like. For example, the machine learning model 10 may output information for controlling the automatic or semi-automatic driving of an automobile. In this case, the machine learning model 10 may be trained using information related to operations performed when a user manually drives the vehicle (e.g., information related to accelerator operation, braking operation, steering operation, etc.). Note that the specific learning algorithm of the machine learning model 10 is not particularly limited, and learning may be performed by appropriately adopting, for example, existing technology.

[0013] The processing device 20 is a device equipped with a processor and a memory. The processor is configured to be able to execute various types of arithmetic processing. The memory is configured to be able to store various types of information. A computer program may be stored in the memory. The processor, together with the memory storing the computer program, may execute various types of processing to be performed by the learning system 1. For example, when the processor executes the computer program, logical functional blocks for executing the processing to be performed by the learning system 1 may be realized within the processor.

[0014] The processing device 20 may be configured to include an input device and an output device in addition to the above-mentioned processor and memory. Specifically, the processing device 20 may include input devices such as a mouse, a keyboard, a touch panel, a microphone, etc. The processing device 20 may also include output devices such as a display, a speaker, etc.

[0015] The processing device 20 includes a learning unit 201, a restriction unit 202, and an analysis unit 203 as functional blocks for realizing its functions. The learning unit 201 is a specific example of a "learning means." The restriction unit 202 is a specific example of a "restriction means." Each of the learning unit 201, the restriction unit 202, and the analysis unit 203 may be realized by a processor or the like included in the processing device 20.

[0016] The learning unit 201 is configured to be able to execute a learning process for the machine learning model 10. The learning unit 201 acquires data from a user and uses the acquired data (i.e., the acquired data is used as learning data) to learn the machine learning model 10. The learning unit 201 may acquire the data used for learning using various sensors. Alternatively, the learning unit 201 may acquire the data used for learning based on user input. Alternatively, the learning unit 201 may acquire the data used for learning from an image of the user.

[0017] As a result of learning by the learning unit 201, the machine learning model 10 is trained to be personalized for the user. If there are multiple users, the learning unit 201 may perform learning for each user. In this case, the machine learning model 10 may be trained for each user (i.e., for each user).

[0018] The limiting unit 202 is configured to be able to limit the output of the machine learning model 10 trained by the learning unit 201. Specifically, the limiting unit 202 limits the output of the machine learning model 10 based on a comparison result between the output of the machine learning model 10 and the output of a standard model. The standard model here is a model that provides a standard output that is prepared in advance. For example, the standard model may be a model that outputs a theoretical standard value. Alternatively, the standard model may be a model that is generated to output the average value of the outputs of multiple models.

[0019] The limiting unit 202 may limit the output of the machine learning model 10, for example, when the difference between the output of the machine learning model 10 and the output of the standard model is equal to or greater than a predetermined value. In other words, the limiting unit 202 may not limit the output of the machine learning model 10 when the difference between the output of the machine learning model 10 and the output of the standard model is not equal to or greater than a predetermined value. The predetermined value here is a threshold value for determining whether or not to limit the output of the machine model 10, and may be set to an appropriate value based on, for example, a prior simulation.

[0020] The analysis unit 203 is configured to be able to perform analysis on the user. Specifically, the analysis unit 203 performs analysis on the user using the difference between the output of the machine learning model 10 and the output of the standard model. The analysis unit 203 may perform analysis on the user using the difference calculated by the restriction unit 202. Note that if the machine learning model 10 outputs multiple parameters, the analysis unit 203 may perform analysis on each of the multiple parameters.

[0021] The analysis unit 203 may analyze how far the user deviates from the standard value depending on the magnitude of the difference. Specifically, the analysis unit 203 may analyze that the user is close to the standard when the difference is small, and may analyze that the user is far from the standard when the difference is large. For example, when the machine learning model 10 is used for driving assistance, the analysis unit 203 may analyze the user's driving operation tendencies (e.g., what kind of driving operation the user prefers, etc.).

[0022] (Operation flow) Next, the flow of operation of the learning system 1 according to the first embodiment (specifically, the operation when limiting the output of the machine learning model 10) will be described with reference to Figure 2. Figure 2 is a flowchart showing the flow of operation of the learning system according to the first embodiment.

[0023] 2, when the operation of the learning system 1 according to the first embodiment starts, the learning unit 201 first executes a learning process for the machine learning model 10 using data acquired from a user (step S11). Note that the learning process by the learning unit 201 may be performed in advance at a timing different from the other processes shown in FIG.

[0024] Next, the limiting unit 202 calculates the difference between the output of the trained machine learning model 10 and the output of the standard model (step S12).Then, the limiting unit 202 determines whether the calculated difference is equal to or greater than a predetermined value (step S13).

[0025] If the difference is greater than or equal to a predetermined value (step S13: YES), the limiting unit 202 limits the output of the machine learning model 10 (step S14). The limiting unit 202 may limit the parameters after the machine learning model 10 outputs the parameters. For example, the limiting unit 202 may execute a process to reduce the parameters output from the machine learning model 10. Alternatively, the limiting unit 202 may limit the machine learning model 10 itself before the machine learning model 10 outputs the parameters. For example, the limiting unit 202 may change the learning parameters of the machine learning model 10 to limit it so that it outputs small parameters from the beginning. Note that if the difference is not greater than or equal to the predetermined value (step S13: NO), the processing of step S14 described above is omitted.

[0026] Next, the analysis unit 203 performs an analysis of the user using the difference between the output of the machine learning model 10 and the output of the standard model (step S15). The analysis unit 203 may output the analysis result to the user. For example, the analysis unit 203 may display the analysis result using a display or the like.

[0027] (Example of output limit) Next, the output restriction of the machine learning model 10 by the learning system 1 according to the first embodiment will be specifically described with reference to Fig. 3. Fig. 3 is a radar chart showing an example of the output restriction of the machine learning model by the learning system according to the first embodiment.

[0028] 3, the machine learning model 10 is configured as a model that outputs five types of parameters, items 1 to 5. Each item may be, for example, an item related to driving assistance. For example, item 1 may be an item related to vehicle acceleration, item 2 may be an item related to vehicle deceleration, item 3 may be an item related to handling, item 4 may be an item related to inter-vehicle distance, and item 5 may be an item related to average speed.

[0029] In the example shown in FIG. 3, item 1 significantly exceeds the chart. Therefore, for item 1, the difference between the output of machine learning model 10 and the output of the standard model is equal to or greater than a predetermined value. Therefore, the limiting unit 202 limits the output of machine learning model 10 related to item 1. Note that the limiting unit 202 does not need to limit the output of the other items 2 to 4.

[0030] The limiting unit 202 may limit a predetermined item among the outputs of multiple items in the machine learning model 10. In other words, the output limit imposed by the limiting unit 202 may be limited to some items. For example, when a difference for a predetermined item is equal to or greater than a predetermined value, the limiting unit 202 may limit the output for that item, while when a difference for an item other than the predetermined item is equal to or greater than the predetermined value, the limiting unit 202 may not limit the output for that item.

[0031] The items to be subject to output limitation may be items related to dangerous driving (for example, items related to sudden acceleration, sudden braking, sudden steering, etc.) In other words, items not related to dangerous driving may be excluded from the output limitation.

[0032] (Technical Effects) Next, the technical effects obtained by the learning system 1 according to the first embodiment will be described.

[0033] As described with reference to FIGS. 1 to 3, in the learning system 1 according to the first embodiment, the output of the machine learning model 10 is limited based on the comparison result with the standard model. In this way, the output of the machine learning model 10 can be made appropriate. This can reduce the inconvenience caused by personalizing the machine learning model 10. For example, by learning the characteristics of a user, it is possible to prevent unintentional manipulation of the subconscious (unethical behavior). More specifically, it is possible to prevent the output of the machine learning model 10 from encouraging dangerous driving as a result of learning the characteristics of a user who is a reckless driver.

[0034] This disclosure is not limited to the above-described embodiments, but may be modified as appropriate within the scope of the claims and the gist or concept of the invention as can be read from the entire specification, and learning systems involving such modifications are also included within the technical scope of the present invention. [Explanation of symbols]

[0035] 1. Learning System 10 Machine Learning Models 20 Processing equipment 201 Learning Department 202 Restricted Section 203 Analysis Department

Claims

[Claim 1] A learning means for learning a machine learning model using data acquired from a user to personalize the user; a limiting means for limiting the output of the machine learning model based on a comparison result between the machine learning model and a predetermined standard model; A learning system that includes:

Images