TCSLib

140 Learning Theory — Convex Minimax Separation

140.1 Overview

This module proves the convex-compact minimax theorem by a Hahn–Banach separation route. For each finite sample \(u\) of columns it forms the upper image of the payoff on that sample, separates the constant vector \((v+\varepsilon ,\dots ,v+\varepsilon )\) from it by a continuous linear functional, and normalizes the coordinates of that functional into a mixed column strategy. This yields the finite sublevel intersection property, which combined with compactness of the row set gives the minimax identity \(\inf _{x} \sup _{y} f(x,y) = \sup _{y} \inf _{x} f(x,y)\).

140.2 Declarations

Definition 140.1 Finite upper image of the payoff
#

For sets \(X, Y \subseteq \mathbb {R}\), a payoff \(f : \mathbb {R}\to \mathbb {R}\to \mathbb {R}\), and a finite sample \(u\) of columns from \(Y\), the finite upper image is the set

\[ U(X,f,u) \; =\; \{ \, z : u \to \mathbb {R}\; \mid \; \exists \, x \in X,\ \forall y \in u,\ f(x,y) \lt z_y \, \} \]

of vectors that strictly dominate the payoff vector of some row on the sample \(u\). This set is written \(\texttt{OnlineLearning.finiteUpperImage}\) in Lean.

Lemma 140.2 Upper image is nonempty

If \(X\) is nonempty, then for every finite column sample \(u\) the upper image \(U(X,f,u)\) is nonempty.

Lemma 140.3 Upper image is convex

If \(X\) is convex and \(x \mapsto f(x,y)\) is convex on \(X\) for every \(y \in Y\), then \(U(X,f,u)\) is a convex subset of \(u \to \mathbb {R}\) for every finite column sample \(u\).

Lemma 140.4 Upper image is upward closed
#

If \(z \in U(X,f,u)\) and \(z_y \le z'_y\) for all \(y \in u\), then \(z' \in U(X,f,u)\) as well; that is, the upper image is closed under coordinatewise increase.

Lemma 140.5 Upper image is open

For every finite column sample \(u\), the upper image \(U(X,f,u)\) is an open subset of \(u \to \mathbb {R}\).

Lemma 140.6 Coordinate expansion of a continuous linear functional

Let \(\iota \) be a finite type and \(L\) a continuous linear functional on \(\iota \to \mathbb {R}\). Then for every \(z : \iota \to \mathbb {R}\),

\[ L(z) \; =\; \sum _{i \in \iota } z_i \, L(e_i), \]

where \(e_i\) is the standard basis vector \(\mathrm{Pi.single}\, i\, 1\).

Lemma 140.7 Separating functional has nonpositive coordinates

Assume \(X\) is nonempty and \(L\) is a continuous linear functional on \(u \to \mathbb {R}\) with \(L(z) \lt L(c)\) for every \(z \in U(X,f,u)\), where \(c\) is a fixed vector. Then \(L(e_y) \le 0\) for every coordinate \(y \in u\).

Lemma 140.8 Separating functional is nonzero

Under the same separation hypothesis \(L(z) \lt L(c)\) for all \(z\) in the upper image, with \(X\) nonempty, the functional \(L\) is not the zero functional.

Lemma 140.9 Positive total weight of the separator

Under the same separation hypothesis, with \(X\) nonempty,

\[ 0 \; \lt \; \sum _{y \in u} \bigl(-L(e_y)\bigr). \]

Consequently the negated coordinates can be normalized to a probability vector on the sampled columns.

Assume the convex-compact minimax hypotheses \(\texttt{OnlineLearning.ConvexCompactMinimaxHypotheses}\) for \(X\), \(Y\), \(f\). Then for every \(\varepsilon \gt 0\) and every finite column sample \(u \subseteq Y\), the set

\[ X \cap \bigcap _{y \in u} \bigl(X \cap \{ x \mid f(x,y) \le v + \varepsilon \} \bigr), \qquad v = \sup _{y \in Y} \inf _{x \in X} f(x,y), \]

is nonempty, the inner sets being the sublevel sets \(\texttt{OnlineLearning.minimaxSublevel}\); i.e. some row \(x \in X\) satisfies \(f(x,y) \le v + \varepsilon \) simultaneously for all sampled columns \(y \in u\).

Theorem 140.11 Minimax identity from finite sublevel intersections

Assume the hypotheses \(\texttt{OnlineLearning.ConvexCompactMinimaxHypotheses}\) for \(X\), \(Y\), \(f\), and suppose that for every \(\varepsilon \gt 0\) and every finite \(u \subseteq Y\) the intersection of \(X\) with the sublevel sets \(\texttt{OnlineLearning.minimaxSublevel}\) at level \(v + \varepsilon \) over \(y \in u\) is nonempty, where \(v = \sup _{y \in Y} \inf _{x \in X} f(x,y)\). Then the minimax identity \(\texttt{OnlineLearning.ConvexCompactMinimaxStatement}\) holds for \(X\), \(Y\), \(f\), i.e.

\[ \inf _{x \in X} \sup _{y \in Y} f(x,y) \; =\; \sup _{y \in Y} \inf _{x \in X} f(x,y). \]
Theorem 140.12 Convex-compact minimax theorem by separation

Under the convex-compact minimax hypotheses \(\texttt{OnlineLearning.ConvexCompactMinimaxHypotheses}\) for \(X\), \(Y\), \(f\), the minimax identity \(\texttt{OnlineLearning.ConvexCompactMinimaxStatement}\) holds:

\[ \inf _{x \in X} \sup _{y \in Y} f(x,y) \; =\; \sup _{y \in Y} \inf _{x \in X} f(x,y). \]

This is the final form of the theorem obtained via the finite-column separation route.