gradientDescent: Gradient descent on weights
In robertness/signalgraph: Signal Processing with Graphs

Gradient descent on weights

1	gradientDescent(wts_init, grad, loss, maxit = 100, epsilon = 0.01)

`loss`	function that takes a weight vector as an input and returns a single value
`maxit`	maximum number of iterations
`epsilon`	stop when change in loss output is less than epsilon 10 iterations in a row.
`initial`	weight values
`gradient`	function that takes weight vector as an input are returns a gradient vector of the same length
`step`	step size