Ting He

Federated learning aggregates model updates from distributed clients, but standard first order methods such as FedAvg apply the same scalar weight to all parameters from each client. Under non-IID data, these uniformly weighted updates can be strongly misaligned across clients, causing client drift and degrading the global model. Here we propose Fisher-Informed Parameterwise Aggregation (FIPA), a second-order aggregation method that replaces client-level scalar weights with parameter-specific Fisher Information Matrix (FIM) weights, enabling true parameter-level scaling that captures how each client's data uniquely influences different parameters. With low-rank approximation, FIPA remains communication- and computation-efficient. Across nonlinear function regression, PDE learning, and image classification, FIPA consistently improves over averaging-based aggregation, and can be effectively combined with state-of-the-art client-side optimization algorithms to further improve image classification accuracy. These results highlight the benefits of FIPA for federated learning under heterogeneous data distributions.

We consider the design of mixing matrices to minimize the operation cost for decentralized federated learning (DFL) in wireless networks, with focus on minimizing the maximum per-node energy consumption. As a critical hyperparameter for DFL, the mixing matrix controls both the convergence rate and the needs of agent-to-agent communications, and has thus been studied extensively. However, existing designs mostly focused on minimizing the communication time, leaving open the minimization of per-node energy consumption that is critical for energy-constrained devices. This work addresses this gap through a theoretically-justified solution for mixing matrix design that aims at minimizing the maximum per-node energy consumption until convergence, while taking into account the broadcast nature of wireless communications. Based on a novel convergence theorem that allows arbitrarily time-varying mixing matrices, we propose a multi-phase design framework that activates time-varying communication topologies under optimized budgets to trade off the per-iteration energy consumption and the convergence rate while balancing the energy consumption across nodes. Our evaluations based on real data have validated the efficacy of the proposed solution in combining the low energy consumption of sparse mixing matrices and the fast convergence of dense mixing matrices.

With the advancement of Artificial Intelligence (AI) towards multiple modalities (language, vision, speech, etc.), multi-modal models have increasingly been used across various applications (e.g., visual question answering or image generation/captioning). Despite the success of AI as a service for multi-modal applications, it relies heavily on clouds, which are constrained by bandwidth, latency, privacy concerns, and unavailability under network or server failures. While on-device AI becomes popular, supporting multiple tasks on edge devices imposes significant resource challenges. To address this, we introduce S2M3, a split-and-share multi-modal architecture for multi-task inference on edge devices. Inspired by the general-purpose nature of multi-modal models, which are composed of multiple modules (encoder, decoder, classifier, etc.), we propose to split multi-modal models at functional-level modules; and then share common modules to reuse them across tasks, thereby reducing resource usage. To address cross-model dependency arising from module sharing, we propose a greedy module-level placement with per-request parallel routing by prioritizing compute-intensive modules. Through experiments on a testbed consisting of 14 multi-modal models across 5 tasks and 10 benchmarks, we demonstrate that S2M3 can reduce memory usage by up to 50% and 62% in single-task and multi-task settings, respectively, without sacrificing accuracy. Furthermore, S2M3 achieves optimal placement in 89 out of 95 instances (93.7%) while reducing inference latency by up to 56.9% on resource-constrained devices, compared to cloud AI.

We revisit a classic problem of inferring the routing tree for a given source in a packet-switched network from end-to-end measurements, with two critical differences from existing solutions: (i) instead of exclusively relying on active measurements obtained by probing, we strive to maximally utilize passive measurements obtained from data packets; (ii) instead of inferring a logical topology that omits degree-2 nodes, we want to recover the physical topology containing all the nodes. Our main idea is to utilize the detailed queueing dynamics inside the network to estimate a certain parameter (residual capacity) of each queue, and then use the estimated parameters as fingerprints to detect the queues shared across paths and thus infer the topology. To this end, we develop a Laplace-transform-based estimator to estimate the parameters of a tandem of queues from end-to-end delays, and efficient algorithms to infer the topology by identifying the parameters associated with the same queue. To improve the accuracy, we further develop a hybrid algorithm that uses the information from active measurements to identify (generalized) siblings and the information from passive measurements to detect shared queues on the paths from the source to each pair of identified siblings. Our inferred topology is guaranteed to converge to the ground-truth topology as the number of measurements increases, up to a permutation of the queues traversed by the same set of paths. Our evaluations in both queueing-theoretic and packet-level simulations show that the proposed solutions, particularly the hybrid algorithm, significantly improve the accuracy over the state of the art.

Overlay network is a non-intrusive mechanism to enhance the existing network infrastructure by building a logical distributed system on top of a physical underlay. A major difficulty in operating overlay networks is the lack of cooperation from the underlay, which is usually under a different network administration. In particular, the lack of knowledge about the underlay topology and link capacities makes the design of efficient overlay routing extremely difficult. In contrast to existing solutions for overlay routing based on simplistic assumptions such as known underlay topology or disjoint routing paths through the underlay, we aim at systematically optimizing overlay routing without causing congestion, by extracting information about the underlay from measurements taken at overlay nodes. To this end, we 1) identify the sufficient information for congestion-free overlay routing, and 2) develop polynomial-complexity algorithms to infer this information with guaranteed accuracy. Our evaluations in NS3 based on real network topologies demonstrate notable performance advantage of the proposed solution over existing solutions.

The decentralized and digitally interconnected future of power grids will be achieved through IT modernization to meet the unprecedented challenges and opportunities amid the rise of distributed energy resources, climate resilience demands, and energy storage advancement.

We consider a novel active learning problem motivated by the need of learning machine learning models for health monitoring in wireless body area network (WBAN). Due to the limited resources at body sensors, collecting each unlabeled sample in WBAN incurs a nontrivial cost. Moreover, training health monitoring models typically requires labels indicating the patient's health state that need to be generated by healthcare professionals, which cannot be obtained at the same pace as data collection. These challenges make our problem fundamentally different from classical active learning, where unlabeled samples are free and labels can be queried in real time. To handle these challenges, we propose a two-phased active learning method, consisting of an online phase where a coreset construction algorithm is proposed to select a subset of unlabeled samples based on their noisy predictions, and an offline phase where the selected samples are labeled to train the target model. The samples selected by our algorithm are proved to yield a guaranteed error in approximating the full dataset in evaluating the loss function. Our evaluation based on real health monitoring data and our own experimentation demonstrates that our solution can drastically save the data curation cost without sacrificing the quality of the target model.