federated learning

Digital Twins in Oncology: From Multimodal Data Integration to Precision Clinical Practice

Oncology digital twins are patient-specific computational models that are built by combining electronic health records, multiomics genomic data, and diagnostic imaging to simulate individual tumor biology and predict multiple treatment-related outcomes. Conceptually originated from aerospace engineering, it has matured clinically through convergent advances in radiomics, mechanistic tumor modeling, pharmacokinetic- pharmacodynamic systems, federated machine learning, and, most recently, large language model (LLM)-based clinical interfaces and agentic artificial intelligence (AI). For a practicing radiologist, digital twins offer a transformative role: imaging-derived quantitative features serve as the primary data, positioning diagnostic imaging at the center of these personalized oncology workflows. Key clinical applications especially in oncology span from chemotherapy response prediction, immunotherapy patient selection, personalized radiation planning, tumor progression modeling, and treatment toxicity forecasting. Several of these applications are achievable with current technology without any significant infrastructure investment. Substantial challenges include imaging data standards, absence of prospective validation, algorithmic bias in underrepresented populations, and regulatory uncertainty for continuously self-updating AI. This narrative review provides radiologists with a balanced, comprehensive overview of digital twin architecture, advanced enabling technologies, current clinical evidence, a practical roadmap for implementation, and a candid appraisal of barriers to adoption.

Information-Theoretic Foundation for Trustworthy Federated Learning

Federated Learning (FL) offers a promising approach to training machine learning models on decentralized data while preserving data privacy. However, the inherent heterogeneity of data across clients and potential malicious behavior introduce challenges to the trustworthiness of FL systems. This work proposes a novel information-theoretic framework for quantifying trust in FL, moving away from traditional subjective trust assumptions. We define a "trust score" based on the mutual information between local model updates and a global consensus model. This score directly measures the contribution of each client's update to the collective knowledge, providing a rigorous metric for assessing model divergence and identifying potential outliers. The core contribution of this paper lies in providing a quantifiable and objective method for evaluating trust in FL, enabling more robust and reliable deployments. We explore the theoretical properties of this mutual information-based trust score and demonstrate its potential for improved outlier detection and model convergence in federated learning scenarios. The framework utilizes concepts from information theory, specifically mutual information, to provide a mathematically grounded approach to assessing the quality and trustworthiness of decentralized model training.

Federated Learning for Cross-Location Data Collaborative Training

This paper presents a novel approach to collaborative model training leveraging federated learning, specifically designed for scenarios involving geographically dispersed data sources. The core challenge addressed is the 'data island' problem, where valuable data remains siloed due to logistical, regulatory, or competitive constraints. Our proposed algorithm, termed Federated Cross-Location Collaborative Training (FCCCT), utilizes the principles of federated learning to enable collaborative model training without direct data sharing. The system operates by iteratively sharing model parameters and gradients between participating data sources. This allows for the construction of a more robust and accurate global model while preserving the privacy of each individual dataset. We demonstrate the effectiveness of FCCCT through a theoretical analysis and outline a practical implementation framework. The key contributions of this work are the integration of federated learning with cross-location data synergy, providing a scalable and privacy-preserving solution for training high-performance models on distributed data. The presented framework offers a significant advancement in addressing the limitations of traditional collaborative learning methods.

Dynamic Differential Privacy for Federated Learning

This paper presents a novel approach to differential privacy in federated learning, termed Dynamic Differential Privacy (DDP). The core idea is to adapt the privacy budget dynamically, responding to the varying sensitivity of local data. Traditional federated learning frequently utilizes a static privacy budget, which can lead to either overly conservative privacy protection, significantly impacting model accuracy, or insufficient protection when confronted with malicious adversaries. DDP addresses these limitations by iteratively adjusting the privacy budget (ε, δ) based on gradient variance and update divergence. The algorithm reduces the privacy budget when gradients exhibit high variance, indicating sensitive data, and increases it when a node provides substantially different updates, signaling potential adversarial behavior. We demonstrate that this dynamic adaptation provides a stronger privacy guarantee while maintaining acceptable accuracy levels compared to fixed-budget approaches. This work contributes to a more robust and practical implementation of federated learning in privacy-sensitive environments.

基于自适应的联邦学习算法

This paper investigates the challenges and opportunities presented by federated learning in dynamic environments. Traditional federated learning approaches often rely on fixed network topology and data distribution, leading to suboptimal model performance and inefficient data utilization. We propose a novel self-adaptive federated learning algorithm that dynamically adjusts model parameters and data sharing strategies based on network topology and data distribution, addressing these limitations. The core mechanism involves a dynamically evolving network structure and a sophisticated adaptive weighting scheme. This framework aims to improve model generalization, reduce communication costs, and enhance data privacy. We present a rigorous theoretical analysis and demonstrate the effectiveness of the proposed algorithm through simulations and experiments. The key innovations lie in the algorithm's ability to continuously re-evaluate the network topology and data distribution, adapting to real-time changes.

Title: Self-Aware Distributed Computing for Algorithmic Discovery

The automation of algorithmic discovery, a critical component of innovation across numerous fields, traditionally relies on the expertise of human researchers and engineers. This paper introduces a novel approach to algorithmic discovery leveraging self-aware distributed computing, specifically employing a federated learning framework combined with a meta-learning algorithm. This architecture aims to foster creativity by creating a collective intelligence capable of generating novel algorithmic solutions through iterative refinement and self-reflection. We propose a distributed system where agents, each possessing a degree of self-awareness, collaboratively explore a space of potential solutions, learning from each other's contributions and adapting their approaches based on feedback. The core mechanism centers around the dynamic generation of new algorithms through a process of meta-learning, where the system learns to refine its own algorithms based on the output of other agents. This paper will detail the proposed architecture, discuss the underlying mathematics, and outline preliminary experimental results demonstrating the potential of this approach to accelerate the discovery of novel algorithms.

Distributed Learning with Federated Bayesian Networks

This paper presents a novel approach to distributed learning leveraging Federated Bayesian Networks (FBNs). Traditional federated learning methods struggle with capturing complex, non-independent data distributions, often leading to suboptimal global models. We introduce a framework where multiple devices collaboratively learn a Bayesian Network model through decentralized updates, explicitly addressing these dependency challenges. The core idea is to allow devices to share probabilistic models and inference results, creating a system that adapts to local data characteristics while maintaining data privacy. The framework incorporates techniques for managing network topology, handling heterogeneous data distributions, and ensuring convergence of the collaborative learning process. We demonstrate the potential of this approach through a theoretical analysis and outline key considerations for practical implementation, highlighting its advantages over conventional federated learning systems. The primary contribution lies in the systematic application of Bayesian Networks to federated learning, providing a robust mechanism for learning from correlated data sources.

Title: Self-Aware Distributed Computing for Algorithmic Discovery

The automation of algorithmic discovery, a critical component of innovation across numerous fields, traditionally relies on the expertise of human researchers and engineers. This paper introduces a novel approach to algorithmic discovery leveraging self-aware distributed computing, specifically employing a federated learning framework combined with a meta-learning algorithm. This architecture aims to foster creativity by creating a collective intelligence capable of generating novel algorithmic solutions through iterative refinement and self-reflection. We propose a distributed system where agents, each possessing a degree of self-awareness, collaboratively explore a space of potential solutions, learning from each other's contributions and adapting their approaches based on feedback. The core mechanism centers around the dynamic generation of new algorithms through a process of meta-learning, where the system learns to refine its own algorithms based on the output of other agents. This paper will detail the proposed architecture, discuss the underlying mathematics, and outline preliminary experimental results demonstrating the potential of this approach to accelerate the discovery of novel algorithms.

An Intelligent Framework for Secure and Scalable Child Abuse Data Management Using Machine Learning and Blockchain

Child abuse remains a critical global concern requiring efficient, secure, and scalable data management systems to support early detection and intervention. However, existing centralized approaches suffer from data fragmentation, limited interoperability, and significant privacy and security challenges. This paper proposes an intelligent hybrid framework that integrates machine learning and blockchain technologies to address these limitations. The framework employs advanced machine learning models for predictive analytics and risk classification, enabling early identification of potential abuse cases from heterogeneous data sources. Simultaneously, blockchain technology is utilized to ensure data integrity, decentralization, and secure access control through immutable ledgers and smart contracts. The system architecture incorporates multi-source data acquisition, preprocessing with anonymization, and off-chain storage mechanisms to enhance scalability while preserving privacy. Experimental results demonstrate that the proposed approach achieves high predictive performance, with the Neural Network model attaining an accuracy of 94.1%, outperforming other baseline models. Furthermore, blockchain integration ensures tamper-proof data management, transparent audit trails, and efficient multi-stakeholder collaboration. The findings highlight the effectiveness of combining machine learning and blockchain in developing secure and intelligent child protection systems. The proposed framework offers a scalable and privacy-preserving solution suitable for real-world deployment across healthcare, social services, and law enforcement domains. Future work will focus on integrating federated learning and optimizing system performance for large-scale applications.

Starfish-FL, an Agentic Federated Learning and Analysis Framework

Starfish-FL is a general-purpose federated learning and analysis platform that orchestrates distributed computation across healthcare institutions without exposing patient-level data. Standardized statistical scripts execute locally behind each institution's firewall and only site-level aggregate outputs are shared for central synthesis. This release adds the r_noninferiority_meta task, which implements federated site-stratified non-inferiority meta-analysis in R. Each site returns a proportion difference and its standard error, and the coordinating centre pools them by inverse-variance weighting using metafor. The release also adds analysis/act-noninferiority/run_analysis.R, a single entry-point script that regenerates Table 1, Table 2 and Table 3 of the accompanying manuscript in one run across the emulated 21-site network and writes the execution log for that run. The individual patient data from the Alteplase Compared to Tenecteplase (AcT) trial are restricted and are not included in this deposit. The script takes the trial extract as an argument and establishes provenance by checking the derived counts against those published by the parent trial, so neither the input path nor the file name is recorded in the deposited log.

MedSparseFL: Sparse and Privacy-Preserving Federated Learning for Healthcare Applications

This project provides processed data indexes for reproducing experiments from the MedSparseFL study, a sparse and privacy-preserving federated learning framework for healthcare applications. The datasets include CheXpert and HAM10000. For each dataset, CSV files (chexpert_index.csv and ham10000_index.csv) contain:Image IDs referencing the official dataset filesLabels used in classification tasksTrain/Validation/Test splits for experiment reproducibility

Traffic road control and lane allocation framework for intelligent transportation system using federated deep reinforcement learning