Name
Title | ||
|---|---|---|
Multiplicative vs. Additive Half-Quadratic Minimization for Robust Cost Optimization. |
Abstract | ||
|---|---|---|
It has been experimentally shown in the literature that half-quadratic (HQ) leads to very effective methods for robust cost minimization when combined with a joint optimization strategy. More precisely, the multiplicative formulation of HQ minimization was employed in these works. In this work we address the questions whether the complementary additive form of HQ is beneficial for solving robust estimation problems. Additive HQ minimization is appealing due to its connection with a quadratic relaxation and because it fully pushes the difficulties induced by robust costs to independent terms in the (lifted) cost function. We also propose a double lifting method combining additive and multiplicative HQ minimization. We report numerical results for synthetic problems and standard bundle adjustment instances. |
Year | Venue | Field |
|---|---|---|
2018 | BMVC | Mathematical optimization,Multiplicative function,Pattern recognition,Computer science,Bundle adjustment,Quadratic equation,Minification,Artificial intelligence |
DocType | Citations | PageRank |
Conference | 2 | 0.36 |
References | Authors | |
0 | 2 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Christopher Zach | 1 | 1457 | 84.01 |
| Guillaume Bourmaud | 2 | 48 | 5.78 |