For a nonempty closed convex subset of a finite-dimensional inner-product space, is the unique point of minimizing .
This closest-point map is the metric projection onto a closed convex set. Repeated projections onto several sets form the separate projections onto convex sets method.
For ,
For a nonempty closed convex set in a Hilbert space, its metric projection onto a closed convex set obeys
This stronger inequality implies nonexpansiveness of metric projection, , by the Cauchy-Schwarz inequality.
Put and . The variational characterization of convex projection gives and . Adding the corresponding inequalities in the common direction yields , which is the displayed estimate.
Thus projection cannot magnify errors in the input, a useful stability property for projected gradient descent and projections onto convex sets. It need not be a strict contraction: projecting two points already in leaves both unchanged. The closedness, convexity and inner-product geometry are essential to this nearest-point theorem.
Let
This closed convex set is nonempty exactly when
These are the minimum and maximum of the linear constraint on the box. If the interval condition fails, there is no projection onto .
For feasible data, minimizing with multiplier for separates across coordinates:
The scalar function is continuous and nonincreasing. On a piece with free coordinates, . Its limiting values are the two box extrema, so one can choose with ; plateaus can make the multiplier nonunique, while the projected vector remains unique by strict convexity. For , feasibility requires and coordinatewise clipping alone suffices.
For example, with , and , taking gives the projection . When is the all-ones vector and is an integer, the constraint set is a capped simplex. The clipped scalar equation permits efficient bisection or a breakpoint search.

Ancestors (5)

  1. Convex set
  2. Mathematical optimization
  3. Area of mathematics
  4. Mathematics
  5. Home

Synonyms (2)

Discussion (0) New discussion

There are no discussions about this article yet.

Articles by others on the same topic (0)

There are currently no matching articles.