Learning apparatus and model learning system
A learning apparatus configured to be able to communicate with a plurality of equipment in which learning models are mounted, wherein when first equipment among the plurality of equipment in which a learning model trained using training data sets acquired in a predetermined first area is mounted and which is controlled by the same is used in a predetermined second area, the learning apparatus uses training data sets acquired in the second area to relearn the learning model mounted in the first equipment.
1. A model learning system comprising a server and a plurality of equipment configured to be able to communicate with the server, wherein
the server determines whether a number of vehicles in a second area is greater than a threshold, and transmits a request for movement from a first area to the second area to first equipment in response to determining that the number of vehicles in the second area is not greater than the threshold,
the server determines whether an amount of training data sets acquired in the second area becomes greater than or equal to a predetermined learning start amount, and
in response to determining that, among the plurality of equipment, the first equipment including a supervised machine learning model learned using training data sets acquired in the first area is used in the second area and determining that the amount of training data sets acquired in the second area becomes greater than or equal to the predetermined learning start amount, the server uses the training data sets acquired in the second area to relearn the supervised machine learning model included in the first equipment,
wherein the training data sets acquired in the second area include measured values of input parameters acquired in the second area and measured values of output parameters corresponding to the measured values of the input parameters,
the first equipment is a self driving vehicle,
when receiving a request for movement from the first area to the second area, the first equipment makes the first equipment move from the first area to the second area based on the request for movement and transmits the supervised machine learning model included in the first equipment to the server and,
when receiving the supervised machine learning model from the first equipment which transmitted the request for movement, the server relearns the received supervised machine learning model using training data sets acquired in the second area and retransmits the relearned supervised machine learning model to the first equipment which received the request for movement.
2. The model learning system according to claim 1 , wherein,
when a number of first equipment making the request for movement is greater than or equal to a predetermined number and the server relearns one or more supervised machine learning models received from the first equipment using training data sets acquired in the second area, the server uses parts of the one or more supervised machine learning models as is and relearns the remaining parts using training data sets acquired in the second area as transfer learning.