In personalized federated learning, each member of a potentially large set of agents aims to train a model minimizing its loss function averaged over its local data distribution. We study this problem under the lens of stochastic optimization, focusing on a scenario with a large number of agents, that each possess very few data samples from their local data distribution. Specifically, we prove novel matching lower and upper bounds on the number of samples required from all agents to approximately minimize the generalization error of a fixed agent. We provide strategies matching these lower bounds, based on a gradient filtering approach: given prior knowledge on some notion of distance between local data distributions, agents filter and aggregate stochastic gradients received from other agents, in order to achieve an optimal bias-variance trade-off. Finally, we quantify the impact of using rough estimations of the distances between local distributions of agents, based on a very small number of local samples.

In personalized federated learning, each member of a potentially large set of agents aims to train a model minimizing its loss function averaged over its local data distribution. We study this problem under the lens of stochastic optimization, focusing on a scenario with a large number of agents, that each possess very few data samples from their local data distribution. Specifically, we prove novel matching lower and upper bounds on the number of samples required from all agents to approximately minimize the generalization error of a fixed agent. We provide strategies matching these lower bounds, based on a gradient filtering approach: given prior knowledge on some notion of distance between local data distributions, agents filter and aggregate stochastic gradients received from other agents, in order to achieve an optimal bias-variance trade-off. Finally, we quantify the impact of using rough estimations of the distances between local distributions of agents, based on a very small number of local samples.

Ranking and selection (R&S) conventionally aims to select the unique best alternative with the largest mean performance from a finite set of alternatives. However, for better supporting decision making, it may be more informative to deliver a small menu of alternatives whose mean performances are among the top $m$. Such problem, called optimal subset selection (OSS), is generally more challenging to address than the conventional R&S. This challenge becomes even more significant when the number of alternatives is considerably large. Thus, the focus of this paper is on addressing the large-scale OSS problem. To achieve this goal, we design a top-$m$ greedy selection mechanism that keeps sampling the current top $m$ alternatives with top $m$ running sample means and propose the explore-first top-$m$ greedy (EFG-$m$) procedure. Through an extended boundary-crossing framework, we prove that the EFG-$m$ procedure is both sample optimal and consistent in terms of the probability of good selection, confirming its effectiveness in solving large-scale OSS problem. Surprisingly, we also demonstrate that the EFG-$m$ procedure enables to achieve an indifference-based ranking within the selected subset of alternatives at no extra cost. This is highly beneficial as it delivers deeper insights to decision-makers, enabling more informed decision-makings. Lastly, numerical experiments validate our results and demonstrate the efficiency of our procedures.