skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.


This content will become publicly available on August 1, 2026

Title: An Efficient Active-Set method With Applications to Sparse Approximations and Risk Minimization
Abstract In this paper we present an efficient active-set method for the solution of convex quadratic programming problems with general piecewise-linear terms in the objective, with applications to sparse approximations and risk-minimization. The algorithm is derived by combining a proximal method of multipliers (PMM) with a standard semismooth Newton method (SSN), and is shown to be globally convergent under minimal assumptions. Further local linear (and potentially superlinear) convergence is shown under standard additional conditions. The major computational bottleneck of the proposed approach arises from the solution of the associated SSN linear systems. These are solved using a Krylov-subspace method, accelerated by certain novel general-purpose preconditioners which are shown to be optimal with respect to the proximal penalty parameters. The preconditioners are easy to store and invert, since they exploit the structure of the nonsmooth terms appearing in the problem’s objective to significantly reduce their memory requirements. We showcase the efficiency, robustness, and scalability of the proposed solver on a variety of problems arising in risk-averse portfolio selection,$$L^1$$ L 1 -regularized partial differential equation constrained optimization, quantile regression, and binary classification via linear support vector machines. We provide computational evidence, on real-world datasets, to demonstrate the ability of the solver to efficiently and competitively handle a diverse set of medium- and large-scale optimization instances.  more » « less
Award ID(s):
2242215
PAR ID:
10621198
Author(s) / Creator(s):
; ;
Publisher / Repository:
Springer
Date Published:
Journal Name:
Journal of Scientific Computing
Volume:
104
Issue:
2
ISSN:
0885-7474
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract In this paper, we present counterexamples to maximal$$L^p$$ L p -regularity for a parabolic PDE. The example is a second-order operator in divergence form with space and time-dependent coefficients. It is well-known from Lions’ theory that such operators admit maximal$$L^2$$ L 2 -regularity on$$H^{-1}$$ H - 1 under a coercivity condition on the coefficients, and without any regularity conditions in time and space. We show that in general one cannot expect maximal$$L^p$$ L p -regularity on$$H^{-1}(\mathbb {R}^d)$$ H - 1 ( R d ) or$$L^2$$ L 2 -regularity on$$L^2(\mathbb {R}^d)$$ L 2 ( R d )
    more » « less
  2. Abstract LetXbe a compact normal complex space of dimensionnandLbe a holomorphic line bundle onX. Suppose that$$\Sigma =(\Sigma _1,\ldots ,\Sigma _\ell )$$ Σ = ( Σ 1 , , Σ ) is an$$\ell $$ -tuple of distinct irreducible proper analytic subsets ofX,$$\tau =(\tau _1,\ldots ,\tau _\ell )$$ τ = ( τ 1 , , τ ) is an$$\ell $$ -tuple of positive real numbers, and let$$H^0_0(X,L^p)$$ H 0 0 ( X , L p ) be the space of holomorphic sections of$$L^p:=L^{\otimes p}$$ L p : = L p that vanish to order at least$$\tau _jp$$ τ j p along$$\Sigma _j$$ Σ j ,$$1\le j\le \ell $$ 1 j . If$$Y\subset X$$ Y X is an irreducible analytic subset of dimensionm, we consider the space$$H^0_0 (X|Y, L^p)$$ H 0 0 ( X | Y , L p ) of holomorphic sections of$$L^p|_Y$$ L p | Y that extend to global holomorphic sections in$$H^0_0(X,L^p)$$ H 0 0 ( X , L p ) . Assuming that the triplet$$(L,\Sigma ,\tau )$$ ( L , Σ , τ ) is big in the sense that$$\dim H^0_0(X,L^p)\sim p^n$$ dim H 0 0 ( X , L p ) p n , we give a general condition onYto ensure that$$\dim H^0_0(X|Y,L^p)\sim p^m$$ dim H 0 0 ( X | Y , L p ) p m . WhenLis endowed with a continuous Hermitian metric, we show that the Fubini-Study currents of the spaces$$H^0_0(X|Y,L^p)$$ H 0 0 ( X | Y , L p ) converge to a certain equilibrium current onY. We apply this to the study of the equidistribution of zeros inYof random holomorphic sections in$$H^0_0(X|Y,L^p)$$ H 0 0 ( X | Y , L p ) as$$p\rightarrow \infty $$ p
    more » « less
  3. Abstract We obtain new optimal estimates for the$$L^{2}(M)\to L^{q}(M)$$ L 2 ( M ) L q ( M ) ,$$q\in (2,q_{c}]$$ q ( 2 , q c ] ,$$q_{c}=2(n+1)/(n-1)$$ q c = 2 ( n + 1 ) / ( n 1 ) , operator norms of spectral projection operators associated with spectral windows$$[\lambda ,\lambda +\delta (\lambda )]$$ [ λ , λ + δ ( λ ) ] , with$$\delta (\lambda )=O((\log \lambda )^{-1})$$ δ ( λ ) = O ( ( log λ ) 1 ) on compact Riemannian manifolds$$(M,g)$$ ( M , g ) of dimension$$n\ge 2$$ n 2 all of whose sectional curvatures are nonpositive or negative. We show that these two different types of estimates are saturated on flat manifolds or manifolds all of whose sectional curvatures are negative. This allows us to classify compact space forms in terms of the size of$$L^{q}$$ L q -norms of quasimodes for each Lebesgue exponent$$q\in (2,q_{c}]$$ q ( 2 , q c ] , even though it is impossible to distinguish between ones of negative or zero curvature sectional curvature for any$$q>q_{c}$$ q > q c
    more » « less
  4. A<sc>bstract</sc> In this paper we explorepp→W±(ℓ±ν)γto$$ \mathcal{O}\left(1/{\Lambda}^4\right) $$ O 1 / Λ 4 in the SMEFT expansion. Calculations to this order are necessary to properly capture SMEFT contributions that grow with energy, as the interference between energy-enhanced SMEFT effects at$$ \mathcal{O}\left(1/{\Lambda}^2\right) $$ O 1 / Λ 2 and the Standard Model is suppressed. We find that there are several dimension eight operators that interfere with the Standard Model and lead to the same energy growth, ~$$ \mathcal{O}\left({E}^4/{\Lambda}^4\right) $$ O E 4 / Λ 4 , as dimension six squared. While energy-enhanced SMEFT contributions are a main focus, our calculation includes the complete set of$$ \mathcal{O}\left(1/{\Lambda}^4\right) $$ O 1 / Λ 4 SMEFT effects consistent with U(3)5flavor symmetry. Additionally, we include the decay of theW±→ ℓ±ν, making the calculation actually$$ \overline{q}{q}^{\prime}\to {\ell}^{\pm}\nu \gamma $$ q ¯ q ± νγ . As such, we are able to study the impact of non-resonant SMEFT operators, such as$$ \left({L}^{\dagger }{\overline{\sigma}}^{\mu }{\tau}^IL\right)\left({Q}^{\dagger }{\overline{\sigma}}^{\nu }{\tau}^IQ\right) $$ L σ ¯ μ τ I L Q σ ¯ ν τ I Q Bμν, which contribute to$$ \overline{q}{q}^{\prime}\to {\ell}^{\pm}\nu \gamma $$ q ¯ q ± νγ directly and not to$$ \overline{q}{q}^{\prime}\to {W}^{\pm}\gamma $$ q ¯ q W ± γ . We show several distributions to illustrate the shape differences of the different contributions. 
    more » « less
  5. Abstract The radiation of steady surface gravity waves by a uniform stream$$U_{0}$$ U 0 over locally confined (width$$L$$ L ) smooth topography is analyzed based on potential flow theory. The linear solution to this classical problem is readily found by Fourier transforms, and the nonlinear response has been studied extensively by numerical methods. Here, an asymptotic analysis is made for subcritical flow$$D/\lambda > 1$$ D / λ > 1 in the low-Froude-number ($$F^{2} \equiv \lambda /L \ll 1$$ F 2 λ / L 1 ) limit, where$$\lambda = U_{0}^{2} /g$$ λ = U 0 2 / g is the lengthscale of radiating gravity waves and$$D$$ D is the uniform water depth. In this regime, the downstream wave amplitude, although formally exponentially small with respect to$$F$$ F , is determined by a fully nonlinear mechanism even for small topography amplitude. It is argued that this mechanism controls the wave response for a broad range of flow conditions, in contrast to linear theory which has very limited validity. 
    more » « less