Hands-On Machine Learning on Google Cloud Platform

Reducing underfitting

Underfitting typically happens when:

• The model is extremely complex and is run for fewer epochs
• The data is not normalized

Scenario 2: Underfitting in action on the MNIST dataset

In the following scenario, we see the case of underfitting in action on the MNIST dataset:

Note that, in the preceding code, we have not scaled our data—the training and test dataset columns have values ranging from 0 to 255:

The TensorBoard ...