Abstract:Recently, several works have shown that natural modifications of the classical conditional gradient method (aka Frank-Wolfe algorithm) for constrained convex optimization, provably converge with a linear rate when: i) the feasible set is a polytope, and ii) the objective is smooth and strongly-convex. However, all of these results suffer from two significant shortcomings: large memory requirement due to the need to store an explicit convex decomposition of the current iterate, and as a consequence, large running-time overhead per iteration, and worst case convergence rate that depends unfavorably on the dimension. In this work we present a new conditional gradient variant and a corresponding analysis that improves on both of the above shortcomings. In particular: both memory and computation overheads are only linear in the dimension. Moreover, in case the optimal solution is sparse, the new convergence rate replaces a factor which is at least linear in the dimension in previous works, with a linear dependence on the number of non-zeros in the optimal solution. At the heart of our method, and corresponding analysis, is a novel way to compute decomposition-invariant away-steps. While our theoretical guarantees do not apply to any polytope, they apply to several important structured polytopes that capture central concepts such as paths in graphs, perfect matchings in bipartite graphs, marginal distributions that arise in structured prediction tasks, and more. Our theoretical findings are complemented by empirical evidence which shows that our method delivers state-of-the-art performance.

Morphological Decomposition of Convex Polytopes and Its Application in Discrete Image Space

Determining Convexo-Concave Vertices of Polygonby Dynamic Distinct Partition

Machining Feature Recognition In Polyhedrons

Generating Large Convex Polytopes Directly on Point Clouds

Decomposing Morphological Structure Element into Neighborhood Configurations

Fast Implementation Of Grayscale Morphological Image Operations By A Hybrid Decomposition Of Structural Function

Incoherent Hybrid Real-Time Morphological Processor

Image Decomposition: Theory, Numerical Schemes, and Performance Evaluation

Differentiable Convex Polyhedra Optimization from Multi-view Images

Equidistant Smoothing of Polyhedra with Arbitrary Topologies

The Polynomial Connection between Morphological Dilation and Discrete Convolution

A splitting primal-dual proximity algorithm for solving composite optimization problems

PolyTO: Structural Topology Optimization using Convex Polygons

Continuous Field Based Free-Form Surface Modeling and Morphing.

Effective Two-Stage Image Segmentation: A New Non-Lipschitz Decomposition Approach with Convergent Algorithm

A HYBRID ALTERNATING MINIMIZATION ALGORITHM FOR STRUCTURED CONVEX OPTIMIZATION PROBLEMS WITH APPLICATION IN POISSONIAN IMAGE PROCESSING

Generalized-Trapezoidal Decomposition Algorithm for Planar Polygons

DECOMPOSING IMAGES INTO TRIANGLES BY DELAUNAY POINT PROCESSES

Linear-memory and Decomposition-invariant Linearly Convergent Conditional Gradient Algorithm for Structured Polytopes

An efficient iterative pseudo point elimination technique to represent the shape of the digital image boundary

Point-in-Convex Polygon and Point-in-Convex Polyhedron Algorithms with O(1) Complexity using Space Subdivision