getGradientFunction: Gradient generating closure
In robertness/signalgraph: Signal Processing with Graphs

Description Usage Details See Also

In the network, any given node v has a set of incoming edges. Those incoming edges have weights. The fitting procedure fits the weights in the network by seperately these weights for each node. More specifically, a seperate optimization is conducted for each node v, where in the weights on the incoming edges to v are optimized on the loss function.

1	getGradientFunction(g, v)

This closure returns a function that computes the gradient of all the weights on the incoming edges of a node v. The output function is intended to be passed to an optimization functional.

gradientDescent, getGradientFunction, doChainRule

robertness/signalgraph documentation built on May 27, 2019, 10:33 a.m.