Fractional hedonic games (FHGs) are an appealing class of hedonic coalition formation games, in which the utility of an agent equals the average utility she ascribes to the members of her coalition. We study the problem of maximizing the social welfare in FHGs both in the offline and online setting. For the offline setting, where the problem is known to be NP-hard, we show that no FPTAS exists. For the adversarial arrival online model, with unrestricted utilities, it is known that no constant competitive ratio is achievable with a deterministic algorithm. Therefore, we investigate two other interesting models of online FHGs. In the first one, the agents arrive in a uniformly random order. For this setting, we provide an asymptotically 1/6-competitive algorithm and prove an upper bound of 1/3 on the possible asymptotic competitive ratio. In the second model, an algorithm may dissolve (at no cost) an existing coalition before assigning an arriving agent to a coalition. Here, we show a 1/(6+4√2)-competitive algorithm and prove an upper bound of 1/(3+2√2) on the possible competitive ratio. For the adversarial arrival model, we show that a randomized algorithm achieves a better competitive ratio for symmetric simple FHGs than any deterministic algorithm. We also provide new results on online matching with random vertex arrival on general graphs, with an asymptotically 1/3-competitive algorithm and an upper bound of 1/3 on the possible asymptotic competitive ratio.
«
Fractional hedonic games (FHGs) are an appealing class of hedonic coalition formation games, in which the utility of an agent equals the average utility she ascribes to the members of her coalition. We study the problem of maximizing the social welfare in FHGs both in the offline and online setting. For the offline setting, where the problem is known to be NP-hard, we show that no FPTAS exists. For the adversarial arrival online model, with unrestricted utilities, it is known that no constant co...
»