MNIST - Reading numbers

Feature Engineering

First we will import and preprocess the dataset. We need to tell the computer how to interpret the text from the dataset as tensors.

Model Designing

Next we design a model to classify the image as showing one of the available numbers. In this case a simple convolution with a dense layer after it to reliably find the class.

Model Training

Using the model architecture we optimize the model weights to produce the lowest loss on the dataset. After each batch the model weights are optimized and once all training samples are processed we begin a new epoch and repeat the process.

Model Prediction

After training we can use the model to classify new numbers.

Model Evaluation

To assess the quality of the model and compare it to others we usually run our model through a suite of tests that give us a feeling about the performance of the model.