Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FCOM: A Federated Collaborative Online Monitoring Framework via Representation Learning

Tanapol Kosolwattana, Huazheng Wang, Raed Al Kontar, Ying Lin

TL;DR

The paper addresses online monitoring of large, heterogeneous populations under resource and privacy constraints. It introduces Federated Collaborative Online Monitoring (FCOM), which learns a shared low-rank representation $Q \in \mathbb{R}^{p\times K}$ and unit-specific coefficients $c_i \in \mathbb{R}^K$ to model per-unit rewards as $f_i(x) = x^T Q c_i$, enabling cross-unit knowledge transfer without data sharing. An alternating least squares–based online federated procedure with event-triggered communication estimates $Q$ and $c_i$; a UCB-based monitoring score guides resource allocation, yielding sublinear regret with a bound $R(T) = O(NK\sqrt{T}\ln T)$ when $K\ll p$, and a controlled communication cost. Empirical results on simulations and online cognitive monitoring in Alzheimer's disease demonstrate improved decision performance and privacy preservation relative to centralized or non-representation-based federated baselines.

Abstract

Online learning has demonstrated notable potential to dynamically allocate limited resources to monitor a large population of processes, effectively balancing the exploitation of processes yielding high rewards, and the exploration of uncertain processes. However, most online learning algorithms were designed under 1) a centralized setting that requires data sharing across processes to obtain an accurate prediction or 2) a homogeneity assumption that estimates a single global model from the decentralized data. To facilitate the online learning of heterogeneous processes from the decentralized data, we propose a federated collaborative online monitoring method, which captures the latent representative models inherent in the population through representation learning and designs a novel federated collaborative UCB algorithm to estimate the representative models from sequentially observed decentralized data. The efficiency of our method is illustrated through theoretical analysis, simulation studies, and decentralized cognitive degradation monitoring in Alzheimer's disease.

FCOM: A Federated Collaborative Online Monitoring Framework via Representation Learning

TL;DR

The paper addresses online monitoring of large, heterogeneous populations under resource and privacy constraints. It introduces Federated Collaborative Online Monitoring (FCOM), which learns a shared low-rank representation and unit-specific coefficients to model per-unit rewards as , enabling cross-unit knowledge transfer without data sharing. An alternating least squares–based online federated procedure with event-triggered communication estimates and ; a UCB-based monitoring score guides resource allocation, yielding sublinear regret with a bound when , and a controlled communication cost. Empirical results on simulations and online cognitive monitoring in Alzheimer's disease demonstrate improved decision performance and privacy preservation relative to centralized or non-representation-based federated baselines.

Abstract

Online learning has demonstrated notable potential to dynamically allocate limited resources to monitor a large population of processes, effectively balancing the exploitation of processes yielding high rewards, and the exploration of uncertain processes. However, most online learning algorithms were designed under 1) a centralized setting that requires data sharing across processes to obtain an accurate prediction or 2) a homogeneity assumption that estimates a single global model from the decentralized data. To facilitate the online learning of heterogeneous processes from the decentralized data, we propose a federated collaborative online monitoring method, which captures the latent representative models inherent in the population through representation learning and designs a novel federated collaborative UCB algorithm to estimate the representative models from sequentially observed decentralized data. The efficiency of our method is illustrated through theoretical analysis, simulation studies, and decentralized cognitive degradation monitoring in Alzheimer's disease.
Paper Structure (15 sections, 3 theorems, 19 equations, 3 figures, 2 tables, 1 algorithm)

This paper contains 15 sections, 3 theorems, 19 equations, 3 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related work
  3. Online monitoring
  4. Federated online learning
  5. method
  6. Problem Statement
  7. Federated Collaborative Online Monitoring
  8. Representation learning-based modeling
  9. FCOM algorithm
  10. Monitoring strategy design
  11. Regret Analysis
  12. Experiments
  13. Simulation Studies
  14. Application to online cognitive monitoring in Alzheimer's disease (AD)
  15. Conclusions & future work

Key Result

Lemma 1

When the Hessian matrices of the objective function in Equations eq:ob_c and eq:ob_q are positive definite at the optimizer $q^*$ and $\Tilde{c^*}$, for any $\epsilon_1 \geq 0$, $\epsilon_2 \geq 0$, $\Vert X_t\Vert_2 \leq S$, $\Vert q_{it}\Vert_2 \leq L$, $\Vert \Tilde{c}_{it}\Vert_2 \leq P$, and fo in which $0 < v_1 < 1$, $0 < v_2 < 1$

Figures (3)

  • Figure 1: a) Illustration of representation learning, b) The overall framework of the proposed Federated Collaborative Online Monitoring (FCOM) algorithm
  • Figure 2: a)-b) The convergence of cumulative regret at $M = 33\%$ and $M = 66\%$, c)-d) Communication cost at $M = 33\%$ and $M = 66\%$
  • Figure 3: The cumulative regret for all bandit algorithms shown in the Alzheimer’s Disease dataset with a) $M = 33\%$ and b) $M = 66\%$

Theorems & Definitions (3)

  • Lemma 1
  • Proposition 1
  • Theorem 1