Hi Kaniska,
In MATLAB, when you train an autoencoder using the trainAutoencoder function from the Deep Learning Toolbox, the default learning rate is not explicitly set by the user in the function call. Instead, it's determined by the training algorithm chosen for the autoencoder. MATLAB uses the scaled conjugate gradient algorithm (trainscg) as the default training function for autoencoders, which adapts its learning parameters dynamically during training, hence not having a fixed "default learning rate" in the traditional sense used in gradient descent-based algorithms.
While for trainAutoencoder and similar functions, you adjust training settings through a different mechanism, typically by setting properties in the autoencoder's training options directly
I hope it helps!
