NonlinearMinimizer

Companion:
class
class Object
trait Matchable
class Any

Type members

Classlikes

case class Projection(proximal: Proximal)
case class ProximalPrimal[T](primal: DiffFunction[T], u: T, z: T, rho: Double)(implicit space: MutableInnerProductModule[T, Double]) extends DiffFunction[T]

Proximal modifications to Primal algorithm for scaled ADMM formulation AdmmObj(x, u, z) = f(x) + rho/2*||x - z + u||2 dAdmmObj/dx = df/dx + rho*(x - z + u)

Proximal modifications to Primal algorithm for scaled ADMM formulation AdmmObj(x, u, z) = f(x) + rho/2*||x - z + u||2 dAdmmObj/dx = df/dx + rho*(x - z + u)

Types

Value members

Concrete methods

def apply(ndim: Int, constraint: Constraint, lambda: Double, usePQN: Boolean): FirstOrderMinimizer[BDV, DiffFunction[BDV]]

A compansion object to generate projection based minimizer that can use SPG/PQN as the solver

A compansion object to generate projection based minimizer that can use SPG/PQN as the solver

Value parameters:
constraint

one of the available constraint, possibilities are x>=0; lb<=x<=ub;aeq*x = beq; 1'x = s, x >= 0; ||x||1 <= s

lambda

the regularization parameter for most of the constraints

ndim

the problem dimension

Returns:

FirstOrderMinimizer to optimize on f(x) and proximal operator

def main(args: Array[String]): Unit
def project(proximal: Proximal, maxIter: Int, m: Int, tolerance: Double, usePQN: Boolean): FirstOrderMinimizer[BDV, DiffFunction[BDV]]

A subset of proximal operators can be represented as Projection operators and for those operators, we give an option to the user to choose a projection based algorithm. The options available for users are SPG (Spectral Projected Gradient) and PQN (Projected Quasi Newton)

A subset of proximal operators can be represented as Projection operators and for those operators, we give an option to the user to choose a projection based algorithm. The options available for users are SPG (Spectral Projected Gradient) and PQN (Projected Quasi Newton)

Value parameters:
proximal

operator that defines proximal algorithm

Returns:

FirstOrderMinimizer to optimize on f(x) and proximal operator