Mercurial > repos > pointsource_algs / file comparison
comparison: src/fb.rs
src/fb.rs
- changeset 4
- 5aa5c279e341
- parent 0
- eb3c7813b67a
equal
deleted
inserted
replaced
| 0:eb3c7813b67a | 4:5aa5c279e341 |
|---|---|
| 455 op𝒟 : &'a 𝒟, | 455 op𝒟 : &'a 𝒟, |
| 456 config : &FBConfig<F>, | 456 config : &FBConfig<F>, |
| 457 iterator : I, | 457 iterator : I, |
| 458 plotter : SeqPlotter<F, N> | 458 plotter : SeqPlotter<F, N> |
| 459 ) -> DiscreteMeasure<Loc<F, N>, F> | 459 ) -> DiscreteMeasure<Loc<F, N>, F> |
| 460 where F : Float + ToNalgebraRealField, | 460 where F : Float + ToNalgebraRealField<MixedType = F>, |
| 461 I : AlgIteratorFactory<IterInfo<F, N>>, | 461 I : AlgIteratorFactory<IterInfo<F, N>>, |
| 462 for<'b> &'b A::Observable : std::ops::Neg<Output=A::Observable>, | 462 for<'b> &'b A::Observable : std::ops::Neg<Output=A::Observable>, |
| 463 //+ std::ops::Mul<F, Output=A::Observable>, <-- FIXME: compiler overflow | 463 //+ std::ops::Mul<F, Output=A::Observable>, <-- FIXME: compiler overflow |
| 464 A::Observable : std::ops::MulAssign<F>, | 464 A::Observable : std::ops::MulAssign<F>, |
| 465 GA : SupportGenerator<F, N, SupportType = S, Id = usize> + Clone, | 465 GA : SupportGenerator<F, N, SupportType = S, Id = usize> + Clone, |
| 519 iterator : I, | 519 iterator : I, |
| 520 mut plotter : SeqPlotter<F, N>, | 520 mut plotter : SeqPlotter<F, N>, |
| 521 mut residual : A::Observable, | 521 mut residual : A::Observable, |
| 522 mut specialisation : Spec, | 522 mut specialisation : Spec, |
| 523 ) -> DiscreteMeasure<Loc<F, N>, F> | 523 ) -> DiscreteMeasure<Loc<F, N>, F> |
| 524 where F : Float + ToNalgebraRealField, | 524 where F : Float + ToNalgebraRealField<MixedType=F>, |
| 525 I : AlgIteratorFactory<IterInfo<F, N>>, | 525 I : AlgIteratorFactory<IterInfo<F, N>>, |
| 526 Spec : FBSpecialisation<F, A::Observable, N>, | 526 Spec : FBSpecialisation<F, A::Observable, N>, |
| 527 A::Observable : std::ops::MulAssign<F>, | 527 A::Observable : std::ops::MulAssign<F>, |
| 528 GA : SupportGenerator<F, N, SupportType = S, Id = usize> + Clone, | 528 GA : SupportGenerator<F, N, SupportType = S, Id = usize> + Clone, |
| 529 A : ForwardModel<Loc<F, N>, F, PreadjointCodomain = BTFN<F, GA, BTA, N>> | 529 A : ForwardModel<Loc<F, N>, F, PreadjointCodomain = BTFN<F, GA, BTA, N>> |
| 655 let à = op𝒟.findim_matrix(μ.iter_locations()); | 655 let à = op𝒟.findim_matrix(μ.iter_locations()); |
| 656 let g̃ = DVector::from_iterator(μ.len(), | 656 let g̃ = DVector::from_iterator(μ.len(), |
| 657 μ.iter_locations() | 657 μ.iter_locations() |
| 658 .map(|ζ| minus_τv.apply(ζ) + ω0.apply(ζ)) | 658 .map(|ζ| minus_τv.apply(ζ) + ω0.apply(ζ)) |
| 659 .map(F::to_nalgebra_mixed)); | 659 .map(F::to_nalgebra_mixed)); |
| 660 let mut x = μ.masses_dvector(); | 660 let mut x = μ.masses_mut(); |
| 661 | 661 |
| 662 // The gradient of the forward component of the inner objective is C^*𝒟Cx - g̃. | 662 // The gradient of the forward component of the inner objective is C^*𝒟Cx - g̃. |
| 663 // We have |C^*𝒟Cx|_2 = sup_{|z|_2 ≤ 1} ⟨z, C^*𝒟Cx⟩ = sup_{|z|_2 ≤ 1} ⟨Cz|𝒟Cx⟩ | 663 // We have |C^*𝒟Cx|_2 = sup_{|z|_2 ≤ 1} ⟨z, C^*𝒟Cx⟩ = sup_{|z|_2 ≤ 1} ⟨Cz|𝒟Cx⟩ |
| 664 // ≤ sup_{|z|_2 ≤ 1} |Cz|_ℳ |𝒟Cx|_∞ ≤ sup_{|z|_2 ≤ 1} |Cz|_ℳ |𝒟| |Cx|_ℳ | 664 // ≤ sup_{|z|_2 ≤ 1} |Cz|_ℳ |𝒟Cx|_∞ ≤ sup_{|z|_2 ≤ 1} |Cz|_ℳ |𝒟| |Cx|_ℳ |
| 665 // ≤ sup_{|z|_2 ≤ 1} |z|_1 |𝒟| |x|_1 ≤ sup_{|z|_2 ≤ 1} n |z|_2 |𝒟| |x|_2 | 665 // ≤ sup_{|z|_2 ≤ 1} |z|_1 |𝒟| |x|_1 ≤ sup_{|z|_2 ≤ 1} n |z|_2 |𝒟| |x|_2 |
| 670 let inner_it = config.inner.iterator_options.stop_target(inner_tolerance); | 670 let inner_it = config.inner.iterator_options.stop_target(inner_tolerance); |
| 671 inner_iters += quadratic_nonneg(config.inner.method, &Ã, &g̃, τ*α, &mut x, | 671 inner_iters += quadratic_nonneg(config.inner.method, &Ã, &g̃, τ*α, &mut x, |
| 672 inner_τ, inner_it); | 672 inner_τ, inner_it); |
| 673 | 673 |
| 674 // Update masses of μ based on solution of finite-dimensional subproblem. | 674 // Update masses of μ based on solution of finite-dimensional subproblem. |
| 675 μ.set_masses_dvector(&x); | 675 //μ.set_masses_dvector(&x); |
| 676 } | 676 } |
| 677 | 677 |
| 678 // Form d = ω0 - τv - 𝒟μ = -𝒟(μ - μ^k) - τv for checking the proximate optimality | 678 // Form d = ω0 - τv - 𝒟μ = -𝒟(μ - μ^k) - τv for checking the proximate optimality |
| 679 // conditions in the predual space, and finding new points for insertion, if necessary. | 679 // conditions in the predual space, and finding new points for insertion, if necessary. |
| 680 let mut d = &minus_τv + op𝒟.preapply(μ_diff(&μ, &μ_base)); | 680 let mut d = &minus_τv + op𝒟.preapply(μ_diff(&μ, &μ_base)); |
