A learning curve shows the generalization performance — the performance only on testing data, plotted against the amount of training data used. A fitting graph shows the generalization performance as well as the performance on the training data, but plotted against model complexity.
