Explain how the use of convolutional neural networks (CNNs) in this specific scenario relates to the concept of strong prior probability distributions on model parameters.
The use of CNNs introduces an infinitely strong prior over its weights, which says that weights for one hidden unit must be identical to its neighbor’s, but shifted in space. This allows the model to converge faster during training by restricting the search space of possible weight values, but at the cost of reduced flexibility […]