Loss: indicates the loss value of the training set
Accuracy: Accuracy of training set
Val_loss: indicates the loss value of the test set
Val_accruacy: Accuracy of test sets
The following 5 cases are for reference:
Train loss decreases and test loss decreases, indicating that the network is still learning. (The best)
Train Loss decreases continuously, while test Loss tends to remain unchanged, indicating that the network is over-fitting. (Max Pool or regularization)
Train Loss tends to remain unchanged, while Test Loss keeps decreasing, indicating that the data set is 100% problematic. (Check the dataset)
Train loss tends to remain unchanged while test Loss tends to remain unchanged, indicating that learning has encountered a bottleneck and the learning rate or batch number needs to be reduced. (Reduce learning rate)
Train loss and test loss keep rising, indicating improper network structure design, improper training super-parameter setting, data set cleaning and other problems. (Worst case)
In this case, loss decreases while VAL_Loss tends to remain unchanged, indicating that the network is in the over-fitting state