Optimization, slide 34

Moderator: Statistisches Maschinelles Lernen

NonStop
Mausschubser
Mausschubser
Beiträge: 73
Registriert: 18. Apr 2015 19:15

Optimization, slide 34

Beitrag von NonStop »

On the slide 34 for optimization are showed the constraint minimization problem and corresponding Lagrangian.
Karush-Kuhn-Tucker conditions says, that for inequality constraints, Lagrangian multipliers should always be greater or equal to 0.
Is the Lagrangian on slide 34 not unbounded then? Should the second plus not be a minus?
Zuletzt geändert von NonStop am 14. Jul 2018 16:06, insgesamt 1-mal geändert.

mProg
BASIC-Programmierer
BASIC-Programmierer
Beiträge: 147
Registriert: 25. Apr 2015 00:10

Re: Optimization, slide 34

Beitrag von mProg »

I would also think that the stated optimization problem in slide 38 is unbounded. Either there must be a minus in front of the Lagrange multiplier of the inequality constraint, or the inequality constraint must be a less equal operator i.e. g<=0.

Antworten

Zurück zu „Statistisches Maschinelles Lernen“