1 paper · 1 filter
Patrick Schnell, Nils Thuerey
Of all the vector fields surrounding the minima of recurrent learning setups, the gradient field with its exploding and vanishing updates appears a poor choice for optimization, of…