Premade estimators are similar to the methods available in packages such as scikit-learn, where the input features and output labels are specified, along with the various hyperparameters. A method can then optimize for solving a loss function that is predefined to a default value but can be varied by passing a different function in a parameter.
Let's explore building the training and test datasets in the code:
- Import the relevant packages:
- Import the dataset. We will work on the MNIST dataset for this exercise:
The shapes of images and labels are as follows:
The premade function works on label value instead of ...
