Practical Methods of Optimization
Citations
19,985 citations
15,696 citations
Cites background from "Practical Methods of Optimization"
...This particular dual formulation of the problem is called the Wolfe dual (Fletcher, 1987)....
[...]
...This is a property of any convex programming problem (Fletcher, 1987)....
[...]
...For the primal problem above, the KKT conditions may be stated (Fletcher, 1987): ∂ ∂wν LP = wν − ∑ i αiyixiν = 0 ν = 1, · · · , d (17) ∂ ∂b LP = − ∑ i αiyi = 0 (18) yi(xi · w+ b) − 1 ≥ 0 i = 1, · · · , l (19) αi ≥ 0 ∀i (20) αi(yi(w · xi + b) − 1) = 0 ∀i (21) The KKT conditions are satisfied at the…...
[...]
...For more on nonlinear programming techniques see (Fletcher, 1987; Mangasarian, 1969; McCormick, 1983)....
[...]
...…with any kind of constraints, provided that the intersection of the set of feasible directions with the set of descent directions coincides with the intersection of the set of feasible directions for linearized constraints with the set of descent directions (see Fletcher, 1987; McCormick, 1983))....
[...]
10,696 citations
Cites background from "Practical Methods of Optimization"
...This requirement is made, as we want to ensure the existence and uniqueness (for strict convexity) of a minimum of optimization problems [Fletcher, 1989]....
[...]
...[Fletcher, 1989]....
[...]
8,811 citations
Cites background from "Practical Methods of Optimization"
...Because the matrixassociated with suykens.tex; 24/11/1999; 16:59; p.3 this quadratic programming problem is not indefinite, the soluti n to (11) will be global (Fletcher, 1987)....
[...]
...…(16) One defines the Lagrangian L3(w;b;e;α) = J3(w;b;e) N∑ k=1αkfyk[wTϕ(xk)+b] 1+ekg (17) suykens.tex; 24/11/1999; 16:59; p.4 whereαk are Lagrange multipliers (which can be either positive or negative now due to the equality constraints as follows from the Kuhn-Tucker conditions (Fletcher, 1987))....
[...]
...which is equivalent to ykTwT 0.xk/CbU1 ;k D 1;:::;N; ( 3 )...
[...]
...Furthermore, one can show that hyperplanes ( 3 ) satisfying the constraintkwk2 a have a VC-dimensionh which is bounded by...
[...]
...In order to have the possibility to violate ( 3 ), in case a separating hyperplane in this higher dimensional space does not exist, variables xk are introduced such that ykTwT 0.xk/CbU1 xk ;k D 1;:::;N;...
[...]
7,966 citations
Cites background from "Practical Methods of Optimization"
..., [7, 12]) to improve the proposed step if a trial point has been rejected....
[...]
...Note that if the regularization parameter ζ > 0 is chosen sufficiently small, the optimization problem (30) is the exact penalty formulation [12] of the problem “find the feasible point that is closest (in a weighted norm) to the reference point x̄R ,”...
[...]