Model learning system and model learning device
A model learning system includes a device in which a learning model is used, a data acquisition device that acquires data used for creating a training data set for the learning model, and a model learning device that automatically learns the learning model using the training data set for the learning model created using the data. The model learning device is configured to, when a model accuracy after automatically learning the learning model decreases or does not increase compared to before learning, exclude the learning model from a target to be automatically learned.
1 . A model learning system comprising:
a device in which a learning model is used;
a data acquisition device configured to acquire data used for creating a training data set for the learning model; and
a model learning device comprising:
a processor;
a database; and
a non-transitory memory storing instructions that cause the processor to:
automatically train the learning model using the training data set for the learning model created using the data,
in response to a model accuracy after automatically training the learning model decreases or does not increase compared to before training, exclude the learning model from a target to be automatically trained,
evaluate the model accuracy of the automatically trained learning model after training to generate a result, and
execute the exclusion of the learning model from the target to be automatically trained based on the result, wherein
the database is configured to store an index of accuracy that is configured to measure the model accuracy of the learning model, and information on whether the learning model is subject to retraining, wherein the processor is further caused to, in response to the learning model being excluded from the target to be automatically trained, transmit a model accuracy improvement request for the learning model to an external organization terminal belonging to an external organization, wherein the external organization is an external specialized organization that analyzes a factor that results in a deterioration of accuracy of the learning model and improves the accuracy by using a human expert.
2 . The model learning system according to claim 1 , wherein the model accuracy improvement request includes information on the learning model excluded from the target to be automatically trained, the training data set used for training the learning model, and the data acquisition device that acquired the data used for creating the training data set.
3 . The model learning system according to claim 1 , wherein the model learning device comprises a server.
4 . The model learning system according to claim 3 , further comprising a vehicle that includes the device.
5 . The model learning system according to claim 1 , wherein the processor is caused to prevent re-training of the learning model from being automatically repeated.
6 . A model learning device comprising:
a processor;
a database; and
a non-transitory memory storing instructions that cause the processor to:
automatically train a learning model using a training data set, in response to a model accuracy after automatically training the learning model decreases or does not increase compared to before training, exclude the learning model from a target to be automatically trained,
evaluating the model accuracy of the automatically trained learning model after training to generate a result, and
executing the exclusion of the learning model from the target to be automatically trained based on the result, wherein
the database is configured to store an index of accuracy that is configured to measure the model accuracy of the learning model, and information on whether the learning model is subject to retraining, wherein the processor is further caused to, in response to the learning model being excluded from the target to be automatically trained, transmit a model accuracy improvement request for the learning model to an external organization terminal belonging to an external organization, wherein the external organization is an external specialized organization that analyzes a factor that results in a deterioration of accuracy of the learning model and improves the accuracy by using a human expert.