is it possible to do "weighted" gradient descent on piece wise defined functions?
>inb4 hurr durr retard
like doing a bias of one term over the other, if it is possible, what would the graph/gradients even look like or mean?
its easy to say that the gradients are pointing towards the downward/upward slope, but what would it geometrically look like in R3 for weighted piecewise functions?
I guess the functions would have to be differentiable and smooth. So it would look like any normal gradient, aesthetically. I don't know what you mean by weighted.