With only one hidden layer, the model predicts the prices on the training dataset with a mean error of 0.0071. The figure below shows the relationship between estimated and actual price of the house:
