Skip to main content
. Author manuscript; available in PMC: 2016 Apr 12.
Published in final edited form as: Proc IEEE Int Conf Comput Vis. 2015 Dec;2015:1841–1849. doi: 10.1109/ICCV.2015.214

Algorithm 1 Stochastic coordinate descent on GFN,p

Require: f : GFN,p → ℝ, D ∈ ℝN×N, V0 ∈GFN,p(D)
  1. for t = 1, …, T do

  2. Select rows ℐ ⊆ {1, …,N}

  3. U0VIDII1DI¯IVI¯.
  4. [Q QR] ← U0 for Q nonsingular

  5. Take G ∈ ∂Vℐ·f(V)

  6. G′ ← Gℐ𝒥 + Gℐ𝒥̄RT

  7. Construct a descent curve Y on GFi,p(Dℐℐ) through U0 in the direction of – G′ (22)

  8. Pick step size τt satisfying Armijo-Wolfe condition [18]

  9. Vt+1Y(τt)

  10. end for