Defining and training the CNN model
A CNN will take the input and transform the data into higher dimensions through several layers. Describing artificial intelligence, machine learning, and deep learning models is not within the scope of this book, which focuses on explainable AI.
However, before creating the CNN, let's define the general concepts that determine the type of layers it will contain:
- The convolutional layer applies random filters named kernels to the data; the data is multiplied by weights; the filters are optimized by the CNN weight optimizing function.
- The pooling layer groups features. If you have {1, 1, 1, 1, 1, ..., 1, 1, 1} in an area of the data, you can group them into a smaller representation such as {1, 1, 1}. You still know that a key feature of the image is {1}.
- The dropout layer literally drops some data out. If you have a blue sky with millions of pixels, you can easily take 50% of those pixels out and still know that the...
