What is the meaning of the graph displayed during Model learning in NVDIA Digits?

Loss is a number of how far away it is from the correct answer. The closer it is to 0, the closer it is to the correct answer.
Accuracy is closer to 100% accuracy.
where (train) is the learning value, where (val) is the validation value, and
The data is divided into those used only for learning and those used only for validation, or testing.If loss(train) is down and the value of (accuracy) does not improve, you may be overlearning.