An example that illustrates the basic procedures that Tensorflow 2.x deep learning process. It uses the MNIST built-in image data set. The dense MLP is defined with Tensorflow/Keras functional API. The model is generated and compiled with Adam optimizer, the learning neural network is visualized with plot_model method. The model trained with fit() function, model is evaluated with sparse_categorical_accuracy metric. Meanwhile, the loss function is also evaluated with validation data set. Model saving and loading are also demonstrated with prediction comparison.