Affine images and preimages of convex sets are convex

Affine maps preserve convexity under both images and inverse images
Affine images and preimages of convex sets are convex

Proposition. Let B:XYB:X\to Y be an .

  1. If ΩX\Omega\subset X is , then B(Ω)YB(\Omega)\subset Y is convex.
  2. If ΘY\Theta\subset Y is convex, then the preimage B1(Θ)={xX:B(x)Θ}B^{-1}(\Theta)=\{x\in X:B(x)\in\Theta\} is convex in XX.

Context. This is the main mechanism for generating new convex sets from old ones: apply an affine change of coordinates, or pull back convex constraints.

Proof sketch. For (1), take u=B(x)u=B(x) and v=B(y)v=B(y) with x,yΩx,y\in\Omega and use the defining identity for affine maps:

λu+(1λ)v=B(λx+(1λ)y)B(Ω). \lambda u+(1-\lambda)v = B(\lambda x+(1-\lambda)y)\in B(\Omega).

For (2), if x,yB1(Θ)x,y\in B^{-1}(\Theta) then B(x),B(y)ΘB(x),B(y)\in\Theta, and convexity of Θ\Theta plus the same identity gives B(λx+(1λ)y)ΘB(\lambda x+(1-\lambda)y)\in\Theta.