Weights in Transpose Convolutions

This may be obvious to some, but…
Are the weights that we used in the lectures and those used in the exam just examples?
And when we use this in our network; Are they fitted parameters that we initialize?


Hi David,

Good question.

If the model is to be trained, they are parameters that are initialized and subsequently fitted. If the model is used in transfer learning with downloaded weights, they have a value resulting from training the original model.

Thank you Reinoud!
I appreciate the explanation.