LBFGSIterationData

Value parameters

alpha

Product between rho and the dot product between s and q of this iteration and next iteration, respectively. This value is used to recalculate the q values for past iterations that are kept by the algorithm. For the ''k''-th iteration of the algorithm, the resulting α,,k,, will be: α,,k,, = ρ,,k,, * (s,,k,,^t^ • q,,k+1,,).

s

VectorD containing the difference between the estimates of the variable values (x), each stored in a VectorD of the last 2 iterations of the algorithm. For the ''k''-th iteration of the algorithm, the resulting s,,k,, will be: s,,k,, = x,,k+1,, - x,,k,,.

y

VectorD containing the difference between the gradient vectors (g), each stored in a VectorD of the last 2 iterations of the algorithm. For the ''k''-th iteration of the algorithm, the resulting y,,k,, will be: y,,k,, = g,,k+1,, - g,,k,,.

ys

Dot product between y and s. This value is used in multiple steps when determining the search direction to take when minimizing the function value, hence it is calculated once for each iteration. For the ''k''-th iteration of the algorithm, the resulting ys,,k,, will be: y,,k,,^t^ • s,,k,,.

Attributes

Graph

Reset zoomHide graphShow graph

Supertypes

trait Serializable

trait Product

trait Equals

class Object

trait Matchable

class Any

Show all