Difference between soft and hard weights?

What are soft and hard weight in machine learning attention? I know that wieghts are adjusted by optimizer in deep learning models, as they soft? Also when we import keras layer from keras hub we freeze layers, so are those frozen layer hard weights?

Hi @tbhaxor, Yes you are correct, Soft weights can change during each runtime, whereas hard weights are pre-trained and fine-tuned and remain frozen afterwards, and do not change during training. Thank You.