In the context of deep learning, what is "vanishing gradient problem"?

• A) A problem where gradients become excessively large during training
• B) A problem where gradients become too small to affect model weights during backpropagation
• C) A problem where gradients are inconsistently calculated due to numerical instability
• D) A problem where gradients are not calculated due to a broken neural network architecture