Understanding convolutional networks

CNNs are more advanced neural networks specialized for machine learning with images. Unlike the hidden layers we used before, CNNs have some layers that are not fully connected. These convolutional layers have depth in addition to just width and height. The general principle is that an image is analyzed patch by patch. We can visualize the 7x7 patch in the image as follows:

This reflects a 32x32 greyscale image, with a 7x7 patch. Example of sliding the patch from left to right is given as follows:

If this ...

Get Machine Learning with TensorFlow 1.x now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.