Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Generalizable Hyperparameter Optimization for Federated Learning on Non-IID Cancer Images

Elisa Gonçalves Ribeiro, Rodrigo Moreira, Larissa Ferreira Rodrigues Moreira, André Ricardo Backes

TL;DR

The paper investigates whether hyperparameters tuned via centralized Bayesian optimization on one cancer histopathology dataset can generalize to non-IID federated learning across tumor types. It introduces a simple cross-dataset aggregation heuristic that averages learning rates and uses modal optimizers and batch sizes to form a combined hyperparameter configuration, alongside dataset-specific optima. Evaluated across six CNNs on ovarian and colon cancer tasks, the combined configuration often yields robust federated performance, with mean F1-scores around 0.909 and resilience across non-IID partitions. This work provides a practical, data-efficient approach to generalizable hyperparameter optimization in privacy-preserving federated medical imaging settings.

Abstract

Deep learning for cancer histopathology training conflicts with privacy constraints in clinical settings. Federated Learning (FL) mitigates this by keeping data local; however, its performance depends on hyperparameter choices under non-independent and identically distributed (non-IID) client datasets. This paper examined whether hyperparameters optimized on one cancer imaging dataset generalized across non-IID federated scenarios. We considered binary histopathology tasks for ovarian and colorectal cancers. We perform centralized Bayesian hyperparameter optimization and transfer dataset-specific optima to the non-IID FL setup. The main contribution of this study is the introduction of a simple cross-dataset aggregation heuristic by combining configurations by averaging the learning rates and considering the modal optimizers and batch sizes. This combined configuration achieves a competitive classification performance.

Generalizable Hyperparameter Optimization for Federated Learning on Non-IID Cancer Images

TL;DR

The paper investigates whether hyperparameters tuned via centralized Bayesian optimization on one cancer histopathology dataset can generalize to non-IID federated learning across tumor types. It introduces a simple cross-dataset aggregation heuristic that averages learning rates and uses modal optimizers and batch sizes to form a combined hyperparameter configuration, alongside dataset-specific optima. Evaluated across six CNNs on ovarian and colon cancer tasks, the combined configuration often yields robust federated performance, with mean F1-scores around 0.909 and resilience across non-IID partitions. This work provides a practical, data-efficient approach to generalizable hyperparameter optimization in privacy-preserving federated medical imaging settings.

Abstract

Deep learning for cancer histopathology training conflicts with privacy constraints in clinical settings. Federated Learning (FL) mitigates this by keeping data local; however, its performance depends on hyperparameter choices under non-independent and identically distributed (non-IID) client datasets. This paper examined whether hyperparameters optimized on one cancer imaging dataset generalized across non-IID federated scenarios. We considered binary histopathology tasks for ovarian and colorectal cancers. We perform centralized Bayesian hyperparameter optimization and transfer dataset-specific optima to the non-IID FL setup. The main contribution of this study is the introduction of a simple cross-dataset aggregation heuristic by combining configurations by averaging the learning rates and considering the modal optimizers and batch sizes. This combined configuration achieves a competitive classification performance.
Paper Structure (12 sections, 8 equations, 3 figures, 5 tables, 1 algorithm)

This paper contains 12 sections, 8 equations, 3 figures, 5 tables, 1 algorithm.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. MATERIAL AND METHODS
  4. Medical Imaging Datasets
  5. CNN Architectures
  6. Federated Learning Pipeline
  7. Hyperparameter Optimization
  8. Combined Hyperparameter Aggregation Heuristic
  9. RESULTS AND DISCUSSION
  10. Centralized HPO
  11. Federated Performance Under Different Hyperparameter Schemes
  12. CONCLUSION

Figures (3)

  • Figure 1: Proposed Method. Phase 1 performs centralized Bayesian hyperparameter optimization with TPE on ovarian and colon datasets, yielding task-specific optima $\lambda^{\star}_{\mathrm{ov}}$ and $\lambda^{\star}_{\mathrm{co}}$ and a combined configuration $\bar{\lambda}$. Phase 2 evaluates these three configurations in a non-IID federated setting using FedAvg.
  • Figure 2: Examples of images from the two datasets used in this study.
  • Figure 3: Federated F1-score for each non-IID scenario evaluated.