Closed ZachariahRosenberg closed 2 years ago
In the VPG implementation, the value loss is calculated,
value_loss = value_error.pow(2).mul(0.5).mean()
Isn't the value loss simply the MSE, so just value_error.pow(2).mean()? Why the additional multiplication of 0.5?
value_error.pow(2).mean()
Thank you!
Is this so that the "2" is removed from the derivative i.e. f(x) = [(value_error) **2] / 2, so therefore f'(x) = value_error ?
f(x) = [(value_error) **2] / 2
f'(x) = value_error
In the VPG implementation, the value loss is calculated,
value_loss = value_error.pow(2).mul(0.5).mean()
Isn't the value loss simply the MSE, so just
value_error.pow(2).mean()
? Why the additional multiplication of 0.5?Thank you!