The weighted arithmetic mean-geometric mean inequality, often abbreviated as weighted AM-GM, stands as a cornerstone of classical analysis and a powerful tool across optimization and probability. For any list of non-negative real numbers, assigning non-negative weights that sum to one produces a bound where the weighted geometric mean never exceeds the weighted arithmetic mean. This generalization of the familiar two-variable or n-variable AM-GM inequalities captures the intuitive idea that averaging with emphasis should produce a value that balances the total, while the geometric mean reflects a multiplicative, scale-invariant center of mass.
Statement and Core Intuition
Consider a finite set of non-negative real numbers \(x_1, x_2, \dots, x_n\) and corresponding weights \(w_1, w_2, \dots, w_n\) that are non-negative and sum to one. The weighted arithmetic mean is defined as \(A = \sum_{i=1}^n w_i x_i\), while the weighted geometric mean is \(G = \prod_{i=1}^n x_i^{w_i}\). The weighted AM-GM inequality asserts that \(G \leq A\), with equality if and only if all positive \(x_i\) corresponding to positive weights are equal. Conceptually, the arithmetic mean linearly aggregates the values, whereas the geometric mean captures their compounded growth rate, and linear aggregation inherently dominates compounding when the weights concentrate mass.
Historical Context and Mathematical Lineage
Although attributed in popular treatments to figures like Cauchy and Jensen, the roots of the weighted inequality stretch into the 18th and 19th centuries through the works of masters such as Euler and the development of convex function theory. The inequality is a direct consequence of Jensen's inequality applied to the concave function logarithm, \(\log(\prod x_i^{w_i}) = \sum w_i \log x_i \leq \log(\sum w_i x_i)\). This connection to convexity reveals why the result is so robust: it applies not only to finite sequences but also to integrals and expectations, making it a fundamental bridge between discrete and continuous analysis.
Proof Techniques and Illustrative Examples
Multiple elegant approaches establish the weighted AM-GM. A common proof uses the standard AM-GM after an exponential change of variables, transforming the weighted geometric mean into an unweighted geometric mean of modified terms. Another route employs Lagrange multipliers to maximize the geometric mean subject to a fixed weighted sum, naturally leading to the equality condition. For example, with weights \((\frac{1}{3}, \frac{2}{3})\) and numbers \(4\) and \(9\), the arithmetic mean is \(\frac{1}{3} \cdot 4 + \frac{2}{3} \cdot 9 = \frac{22}{3}\), while the geometric mean is \(4^{1/3} \cdot 9^{2/3} = \sqrt[3]{324} \approx 6.87\), clearly below the arithmetic mean.
Role in Information Theory and Economics
In information theory, the weighted AM-GM underpins the derivation of channel capacities and the behavior of weighted combinations of distributions, where logarithmic measures like entropy rely on the inequality to establish fundamental limits. Economists use it to model Cobb-Douglas production functions and utility representations, where constant returns to scale and substitutability between factors are elegantly captured by weighted means. The inequality ensures that diversified allocations yield higher geometric growth than concentrating on a single factor, reflecting a profound stability principle.
Analytical Applications and Optimization
Analysts leverage weighted AM-GM to derive sharp bounds in inequalities, estimate integrals, and control norms in functional analysis. In optimization, particularly in geometric programming, posynomials and log-convex objectives are transformed using weighted means to reveal convex structures, enabling efficient numerical solutions. The inequality provides a verification tool: if a proposed lower bound for a complicated expression matches the weighted AM-GM bound under optimal weights, the bound is tight and the optimizer is characterized by equality conditions.
