cr.sparse.opt.SmoothFunction

class cr.sparse.opt.SmoothFunction(func: Callable[[jax.Array], float], grad: Callable[[jax.Array, float], jax.Array], grad_val: Callable[[jax.Array, float], Tuple[float, jax.Array]])

Represents a smooth function

Let op be a variable of type SmoothFunction which represents some smooth function \(f\). Then:

• op.func(x) returns the function value \(f(x)\).

• op.grad(x) returns the gradient of function \(g(x) = \nabla f(x)\).

• op.grad_val(x) returns the pair \((g(x), f(x))\).

Attributes

func	Definition of a smooth function
grad	Definition of a gradient the function
grad_val	A wrapper to evaluate the gradient vector and the function value together