Single Parameter Combinatorial Auctions with Partially Public Valuations

We consider the problem of designing truthful auctions, when the bidders’ valuations have a public and a private component. In particular, we consider combinatorial auctions where the valuation of an agent $i$ for a set $S$ of items can be expressed as $v_if(S)$, where $v_i$ is a private single parameter of the agent, and the function $f$ is publicly known. Our motivation behind studying this problem is two-fold: (a) Such valuation functions arise naturally in the case of ad-slots in broadcast media such as Television and Radio. For an ad shown in a set $S$ of ad-slots, $f(S)$ is, say, the number of {\em unique} viewers reached by the ad, and $v_i$ is the valuation per-unique-viewer. (b) From a theoretical point of view, this factorization of the valuation function simplifies the bidding language, and renders the combinatorial auction more amenable to better approximation factors. We present a general technique, based on maximal-in-range mechanisms, that converts any $\alpha$-approximation non-truthful algorithm ($\alpha \leq 1$) for this problem into $\Omega(\frac{\alpha}{\log{n}})$ and $\Omega(\alpha)$-approximate truthful mechanisms which run in polynomial time and quasi-polynomial time, respectively.

💡 Research Summary

The paper tackles the design of truthful mechanisms for combinatorial auctions in which each bidder’s valuation for a set of items can be written as the product of a private scalar v_i and a publicly known set function f(S). This “public‑private factorization” captures many real‑world scenarios, most notably advertising slots in broadcast media: f(S) may represent the number of unique viewers reached by showing an ad in the slot set S, while v_i is the advertiser’s value per unique viewer. By separating the valuation into a public component that is common to all bidders and a private single‑parameter, the authors obtain a bidding language that is both expressive enough for interesting combinatorial effects and simple enough to admit algorithmic treatment.

The central contribution is a generic transformation that turns any α‑approximation algorithm for the underlying allocation problem (where α ≤ 1) into a truthful mechanism with provable approximation guarantees. The transformation is built on the maximal‑in‑range (MIR) paradigm: one first defines a restricted range R of feasible allocations that is small enough to be searched exhaustively in polynomial (or quasi‑polynomial) time, yet rich enough to contain an allocation whose social welfare is within a constant factor of the optimum produced by the original algorithm. The mechanism then selects the welfare‑maximizing allocation inside R and charges each bidder a VCG‑style payment computed with respect to the restricted range. Because the payment rule depends only on the public function f and the chosen allocation, bidders cannot benefit from misreporting their private scalar v_i, guaranteeing truthfulness.

Two concrete instantiations of the framework are analyzed. In the polynomial‑time version, the range R is constructed by enumerating a logarithmic‑size family of candidate allocations derived from the input α‑approximation algorithm (e.g., by running the algorithm on a set of scaled bids). This yields a truthful mechanism whose approximation ratio is Ω(α / log n), where n is the number of items. The loss of a log n factor is the price paid for keeping the range small enough for polynomial‑time enumeration. In the quasi‑polynomial‑time version, the authors allow the range to be larger—of size n^{O(log n)}—by a more exhaustive combination of candidate allocations. In this setting the mechanism retains the full α‑approximation factor, i.e., it is Ω(α)‑approximate, while still being computationally feasible (quasi‑polynomial time).

The paper situates these results within the broader literature on combinatorial auctions. Traditional multi‑parameter auctions suffer from an inherent tension between computational tractability and incentive compatibility: optimal VCG mechanisms are computationally intractable for most combinatorial settings, and known truthful approximations often achieve only logarithmic or worse approximation ratios. By exploiting the single‑parameter nature of the private component, the authors bypass many of these obstacles. Their approach can be combined with any existing α‑approximation algorithm for the underlying allocation problem—whether greedy, local‑search, or LP‑based—so the quality of the final truthful mechanism inherits the best known approximation guarantees for the specific class of public functions f. For instance, when f is submodular, known (1‑1/e)‑approximation algorithms can be plugged in, leading to truthful mechanisms with approximation Ω((1‑1/e)/log n) in polynomial time and Ω(1‑1/e) in quasi‑polynomial time.

Beyond the theoretical analysis, the authors discuss several natural applications. In broadcast advertising, the public function f captures audience reach, which is typically measured by Nielsen ratings or digital analytics and is publicly available. Advertisers only need to submit a per‑viewer bid, dramatically simplifying the bidding process. Similar structures appear in cloud computing (where f could be total processing capacity of a set of servers) and in spectrum allocation (where f could be the geographic coverage of a set of frequencies). In each case, the mechanism’s payments are transparent and depend only on publicly observable quantities, enhancing trust and ease of implementation.

The paper also acknowledges limitations and directions for future work. The logarithmic loss in the polynomial‑time version suggests that tighter range constructions or smarter compression of candidate allocations could improve the approximation factor. Moreover, the framework assumes that the public function f is known exactly and can be evaluated efficiently; extending the results to settings with noisy or stochastic estimates of f would broaden applicability. Finally, while the authors provide rigorous worst‑case guarantees, empirical evaluation on real advertising or cloud workload data is left for future research.

In summary, the authors present a powerful and versatile technique that bridges the gap between non‑truthful approximation algorithms and truthful mechanisms for a broad class of combinatorial auctions with partially public valuations. By leveraging maximal‑in‑range mechanisms, they achieve Ω(α/ log n) approximation in polynomial time and Ω(α) approximation in quasi‑polynomial time, thereby advancing both the theory and potential practice of truthful combinatorial auction design.