TCSLib

108 Learning Theory — CCE

108.1 Overview

This module develops the theory of coarse correlated equilibria (CCE) for finite two-player games. It introduces joint distributions over action profiles, defines (approximate) CCE in terms of one-sided deviation utilities, constructs the empirical joint distribution from a sequence of mixed-strategy plays, and proves the main result: if both players run no-regret algorithms with per-round external regret at most \(\varepsilon \), then their empirical joint distribution is an \(\varepsilon \)-CCE.

108.2 Declarations

Definition 108.1 Two-player finite game
#

A finite two-player game with \(M\) row actions and \(N\) column actions, specified by a row utility function \(u_r : \mathrm{Fin}\, M \times \mathrm{Fin}\, N \to \mathbb {R}\) and a column utility function \(u_c : \mathrm{Fin}\, M \times \mathrm{Fin}\, N \to \mathbb {R}\). Both players are utility maximizers.

Definition 108.2 Embedding a zero-sum game
#

Given a zero-sum game \(G\) with payoff matrix \(A\), this embeds it into the general two-player framework by setting the row utility to \(A\) and the column utility to \(-A\), so that both players are maximizers of their respective utilities.

Definition 108.3 Joint distribution over action profiles
#

A joint distribution \(\sigma \) on \(\mathrm{Fin}\, M \times \mathrm{Fin}\, N\) is a nonnegative function \(\sigma _{ij} \ge 0\) satisfying \(\sum _{i,j} \sigma _{ij} = 1\). It represents a correlated distribution over pure action profiles of the two players.

Definition 108.4 Row marginal
#

The row marginal of \(\sigma \) at row action \(i\) is \(\sigma ^{\mathrm{row}}_i = \sum _{j} \sigma _{ij}\), i.e. the probability that the row player plays \(i\) under \(\sigma \).

Definition 108.5 Column marginal
#

The column marginal of \(\sigma \) at column action \(j\) is \(\sigma ^{\mathrm{col}}_j = \sum _{i} \sigma _{ij}\), i.e. the probability that the column player plays \(j\) under \(\sigma \).

Lemma 108.6 Row marginals are nonnegative

For any joint distribution \(\sigma \) and row action \(i\), \(\sigma ^{\mathrm{row}}_i \ge 0\).

Lemma 108.7 Column marginals are nonnegative

For any joint distribution \(\sigma \) and column action \(j\), \(\sigma ^{\mathrm{col}}_j \ge 0\).

Lemma 108.8 Row marginals sum to one

The row marginals form a valid probability distribution: \(\sum _{i} \sigma ^{\mathrm{row}}_i = 1\).

Lemma 108.9 Column marginals sum to one

The column marginals form a valid probability distribution: \(\sum _{j} \sigma ^{\mathrm{col}}_j = 1\).

Definition 108.10 Row expected utility

The row player’s expected utility under \(\sigma \) in game \(G\) is \(\sum _{i,j} \sigma _{ij} \cdot u_r(i,j)\).

Definition 108.11 Column expected utility

The column player’s expected utility under \(\sigma \) in game \(G\) is \(\sum _{i,j} \sigma _{ij} \cdot u_c(i,j)\).

Definition 108.12 Row deviation utility

The row player’s expected utility from unilaterally deviating to the fixed pure action \(i'\), while the column player keeps \(\sigma \)’s column marginal: \(\sum _{j} \sigma ^{\mathrm{col}}_j \cdot u_r(i', j)\).

Definition 108.13 Column deviation utility

The column player’s expected utility from unilaterally deviating to the fixed pure action \(j'\), while the row player keeps \(\sigma \)’s row marginal: \(\sum _{i} \sigma ^{\mathrm{row}}_i \cdot u_c(i, j')\).

Definition 108.14 Coarse correlated equilibrium

A joint distribution \(\sigma \) is a coarse correlated equilibrium (CCE) of game \(G\) if neither player can improve their expected utility by committing in advance to any fixed pure action: for all \(i'\), \(\mathrm{rowDevUtil}(\sigma , G, i') \le \mathrm{rowExpUtil}(\sigma , G)\), and for all \(j'\), \(\mathrm{colDevUtil}(\sigma , G, j') \le \mathrm{colExpUtil}(\sigma , G)\).

Definition 108.15 \(\varepsilon \)-coarse correlated equilibrium

A joint distribution \(\sigma \) is an \(\varepsilon \)-coarse correlated equilibrium (\(\varepsilon \)-CCE) of game \(G\) if each player’s gain from any unilateral pure deviation is at most \(\varepsilon \): for all \(i'\), \(\mathrm{rowDevUtil}(\sigma , G, i') \le \mathrm{rowExpUtil}(\sigma , G) + \varepsilon \), and similarly for the column player.

Lemma 108.16 Exact CCE implies approximate CCE

If \(\sigma \) is an exact CCE of \(G\) and \(\varepsilon \ge 0\), then \(\sigma \) is also an \(\varepsilon \)-CCE of \(G\).

Definition 108.17 Product distribution
#

Given mixed strategies \(p\) for the row player and \(q\) for the column player, the product (independent) joint distribution assigns to each profile \((i,j)\) the probability \(p_i \cdot q_j\).

Lemma 108.18 Product distribution row marginal

For the product distribution of \(p\) and \(q\), the row marginal at \(i\) equals \(p_i\) (the original row mixed-strategy weight).

Lemma 108.19 Product distribution column marginal

For the product distribution of \(p\) and \(q\), the column marginal at \(j\) equals \(q_j\) (the original column mixed-strategy weight).

Definition 108.20 Empirical joint distribution
#

Given \(T \gt 0\) rounds of play in which the row player uses mixed strategy \(p_t\) and the column player uses \(q_t\), the empirical joint distribution assigns to each profile \((i,j)\) the time-averaged probability \(\frac{1}{T}\sum _{t=1}^{T} (p_t)_i \cdot (q_t)_j\).

Lemma 108.21 Empirical joint row marginal

The row marginal of the empirical joint distribution at action \(i\) equals the time-average of the row player’s mixed-strategy weights: \((1/T)\sum _{t} (p_t)_i\).

Lemma 108.22 Empirical joint column marginal

The column marginal of the empirical joint distribution at action \(j\) equals the time-average of the column player’s mixed-strategy weights: \((1/T)\sum _{t} (q_t)_j\).

Lemma 108.23 Empirical utility sum factorization

For any function \(f : \mathrm{Fin}\, M \to \mathrm{Fin}\, N \to \mathbb {R}\), the sum \(\sum _{i,j} \sigma _{ij} \cdot f(i,j)\) over the empirical joint distribution equals the time-average \((1/T)\sum _{t}\sum _{i,j}(p_t)_i(q_t)_j f(i,j)\).

Lemma 108.24 Empirical row expected utility

The row player’s expected utility under the empirical joint distribution equals the time-average of the per-round bilinear utilities:

\[ \mathrm{rowExpUtil}(\sigma ^{\mathrm{emp}}, G) = \frac{1}{T}\sum _{t}\sum _{i,j}(p_t)_i\, (q_t)_j\, u_r(i,j). \]
Lemma 108.25 Empirical column expected utility

The column player’s expected utility under the empirical joint distribution equals the time-average of the per-round bilinear column utilities:

\[ \mathrm{colExpUtil}(\sigma ^{\mathrm{emp}}, G) = \frac{1}{T}\sum _{t}\sum _{i,j}(p_t)_i\, (q_t)_j\, u_c(i,j). \]
Lemma 108.26 Empirical row deviation utility

The row player’s deviation utility for pure action \(i'\) in the empirical distribution equals the time-average utility of always playing \(i'\):

\[ \mathrm{rowDevUtil}(\sigma ^{\mathrm{emp}}, G, i') = \frac{1}{T}\sum _{t}\sum _{j}(q_t)_j\, u_r(i',j). \]
Lemma 108.27 Empirical column deviation utility

The column player’s deviation utility for pure action \(j'\) in the empirical distribution equals the time-average utility of always playing \(j'\):

\[ \mathrm{colDevUtil}(\sigma ^{\mathrm{emp}}, G, j') = \frac{1}{T}\sum _{t}\sum _{i}(p_t)_i\, u_c(i,j'). \]
Theorem 108.28 No-regret play yields an approximate CCE

Let \(G\) be a two-player game, \(T \gt 0\), and let \(p_t\), \(q_t\) be the mixed strategies played at round \(t\) by the row and column players respectively. Suppose that for every pure row deviation \(i'\),

\[ \sum _{t}\sum _{j}(q_t)_j u_r(i',j) - \sum _{t}\sum _{i,j}(p_t)_i(q_t)_j u_r(i,j) \le \varepsilon T, \]

and similarly for every column deviation \(j'\) using \(u_c\). Then the empirical joint distribution is an \(\varepsilon \)-CCE of \(G\).