Difficulty to understand the ResNet

Hello, I don’t understand why resnets are useful actually and how does it help in the improvment in learning a NN, what iam thinking about that it is just like a safety margin mechanism to the vanishing gradient/activation problem such that once my gradients and activation vanish i like have a checkpoint before just to keep my NN still alive and prevent it from being dead.

is this true or i have misunderstood it?
Thanks.

You got it quite right!

As you say you have some back up information before the point the model stopped learning which you can use for your learning process overall and final output.

1 Like