A constraint is a condition that rules out choices or limits what a method can achieve. In optimization, it defines which candidate solutions are permissible. In learning, assumptions, data, representation, and computation bound how much uncertainty can be reduced. section 8.2.3 section 13.4.1
Constraints can limit different resources or conditions:
- A project portfolio may be limited by budgets, staffing, risk, and category requirements; satisfying one limit does not ensure the portfolio satisfies the others. section 6.4.1
- A hard-margin support vector machine requires each training point to lie on or beyond its class’s margin boundary. section 8.2.3
- Scaling learning systems remains bounded by physical resources, reliable information, and economic costs. section 13.5
Portfolio selection
A project can appear attractive under the annual budget yet make a portfolio infeasible under its total budget or category limits. Feasibility depends on satisfying all applicable constraints together. section 6.4.1
In a hard-margin SVM, this inequality requires every labeled training point to lie on or beyond its appropriate margin boundary. section 8.2.3




























