Struct totsu_core::ConePSD

pub struct ConePSD<'a, L: LinAlgEx> { /* private fields */ }

Positive semidefinite cone

\[ {\rm vec}(\mathcal{S}_+^k) = \left\lbrace x \in \mathbb{R}^n,\ n=\frac12 k(k+1) \ \middle|\ {\rm vec}^{-1}(x) \in \mathcal{S}_+^k \right\rbrace \] \( {\rm vec}(X) = (X_{11}\ \sqrt2 X_{12}\ X_{22}\ \sqrt2 X_{13}\ \sqrt2 X_{23}\ X_{33}\ \cdots)^T \) which extracts and scales the upper-triangular part of a symmetric matrix X in column-wise.

Implementations

impl<'a, L: LinAlgEx> ConePSD<'a, L>

pub fn query_worklen(nvars: usize) -> usize

Query of a length of work slice.

Returns the length of work slice that ConePSD::new requires.

• nvars is a number of variables, that is a length of x of ConePSD::proj.

pub fn new(work: &'a mut [L::F], eps_zero: L::F) -> Self

Creates an instance.

Returns the ConePSD instance.

• work slice is used for temporal variables in ConePSD::proj.
• eps_zero should be the same value as crate::solver::SolverParam::eps_zero.

Trait Implementations

impl<'a, L: LinAlgEx> Cone<L> for ConePSD<'a, L>

fn proj(&mut self, _dual_cone: bool, x: &mut L::Sl) -> Result<(), ()>

Calculates \(\Pi_\mathcal{K}(x)\), that is euclidean projection of \(x\) onto the cone \(\mathcal{K}\).

fn product_group<G: Fn(&mut L::Sl) + Copy>(&self, dp_tau: &mut L::Sl, group: G)

Performs grouping for a diagonal preconditioning vector according to the cone \(\mathcal{K}\).