LEARNING DEVICE AND FUEL INJECTION SYSTEM

A learning device changes each of multiple parameters to an increased side and to a decreased side with respect to a reference value set for each of the multiple parameters. The learning device calculates an approximation degree for each combination of the changes of the multiple parameters. The learning device performs update by using the combination of the changes providing the highest approximation degree among the approximation degrees as the combination of the updated reference values when an update end condition is not satisfied. The update end condition is a condition that the approximation degree at the reference values is higher than any of the approximation degrees after the change. The learning device decides the reference values as learning values of the multiple parameters when the update end condition is satisfied.