Hi there! as an extra exercise for chapter 4, I tried the following exercise:
Use SGD to find the weights for the displacement equation for spring harmonic movement
you can find the physical background here
What I encountered, is that using MSE, L1 or Huber loss (combination of both) as loss functions in this case only brought a certain reduction of loss, to a certain point (~30 in my example), but couldn’t find adequate weights to give predictions that are close enough to my validation set.
I was wondering if SGD is the right approach to solve harmonic (sine) equations. could someone please help me solve this?
Here’s a link to my notebook