parameters

Parameters are the ones that are learnt from the data during training process.

They could be as simple as a scalar value like bias in the case of a neural net layer, or a 4-D filter in the case of a convolutional layer. They are typically learnt using iterative methods like SGD, Adam, etc.