Ask HN: Can traditional ML methods improve with feedback loop (a la backprop)?

[bobosha]

Is there any research to implement methods analogous to backprop (in deep learning), in classic ML methods such as SVMs, RF etc?