> so residual error derivatives in a sense? Yep, loss = residuals*2 + lambda*sum(|a_i|), and we take the gradient of that with respect to the a_i to guide our

> so residual error derivatives in a sense? Yep, loss = residuals*2 + lambda*sum(|a_i|), and we take the gradient of that with respect to the a_i to guide our gradient descent steps.