Pasar al contenido principal

APRENDIZAJE FEDERADO Y CONTINUO A PARTIR DE DATOS HETEROGENEOS EN DISPOSITIVOS Y ROBOTS

PID2020-119367RB-I00

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal de I+D+i Orientado a los Retos de la Sociedad
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2020
Plan Estatal 2017-2020 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario UNIVERSIDADE DE SANTIAGO DE COMPOSTELA

Resultados relacionados

Resultados totales (Incluyendo duplicados): 6
Encontrada(s) 1 página(s)

Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

Non-IID data and Continual Learning processes in Federated Learning: a long road ahead

  • Fernández Criado, Marcos
  • Estévez Casado, Fernando
  • Iglesias Rodríguez, Roberto
  • Vázquez Regueiro, Carlos
  • Barro Ameneiro, Senén
Federated Learning is a novel framework that allows multiple devices or institutions to train a machine learning model collaboratively while preserving their data private. This decentralized approach is prone to suffer the consequences of data statistical heterogeneity, both across the different entities and over time, which may lead to a lack of convergence. To avoid such issues, different methods have been proposed in the past few years. However, data may be heterogeneous in lots of different ways, and current proposals do not always determine the kind of heterogeneity they are considering. In this work, we formally classify data statistical heterogeneity and review the most remarkable learning Federated Learning strategies that are able to face it. At the same time, we introduce approaches from other machine learning frameworks. In particular, Continual Learning strategies are worthy of special attention, since they are able to handle habitual kinds of data heterogeneity. Throughout this paper, we present many methods that could be easily adapted to the Federated Learning settings to improve its performance. Apart from theoretically discussing the negative impact of data heterogeneity, we examine it and show some empirical results using different types of non-IID data
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

Ensemble and continual federated learning for classifcation tasks

  • Estévez Casado, Fernando
  • Lema Pais, Dylan
  • Iglesias Rodríguez, Roberto
  • Vázquez Regueiro, Carlos
  • Barro Ameneiro, Senén
Federated learning is the state-of-the-art paradigm for training a learning model collaboratively across multiple distributed devices while ensuring data privacy. Under this framework, different algorithms have been developed in recent years and have been successfully applied to real use cases. The vast majority of work in federated learning assumes static datasets and relies on the use of deep neural networks. However, in real world problems, it is common to have a continual data stream, which may be non stationary, leading to phenomena such as concept drift. Besides, there are many multi-device applications where other, non-deep strategies are more suitable, due to their simplicity, explainability, or generalizability, among other reasons. In this paper we present Ensemble and Continual Federated Learning, a federated architecture based on ensemble techniques for solving continual classification tasks. We propose the global federated model to be an ensemble, consisting of several independent learners, which are locally trained. Thus, we enable a flexible aggregation of heterogeneous client models, which may differ in size, structure, or even algorithmic family. This ensemble-based approach, together with drift detection and adaptation mechanisms, also allows for continual adaptation in situations where data distribution changes over time. In order to test our proposal and illustrate how it works, we have evaluated it in different tasks related to human activity recognition using smartphones
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Analysis of strategies to apply federated learning in the field of robotics

  • Martínez Enríquez, Roi
  • Fraga Corredoira, Nicolás
  • Burés Amatriaín, José Miguel
  • Iglesias Rodríguez, Roberto
  • García Polo, Francisco Javier
  • Fernández Vidal, Xosé Ramón
In this paper we analyse the use of techniques that will help us to apply federated learning in the context of robotics. Federated learning allows the building of a global model from data that is collected locally, at different devices and without data sharing. Nowadays there are already solutions that face the problem of data heterogeneity (non i.i.d). Nevertheless, even when there seem to be a common task that guides the federated learning, the local implementation or details are not exactly identical. This is particularly important in the case of robotics. In this paper we analyse the performance of different techniques that allow us to identify malicious learners, or minority dissenting learners. This identification will allow to modulate the impact of these learners in the global model. Finally, in order to apply federated learning in the context of robotics, we also need strategies which allow the learning of a model when the sensors on the robots are heterogeneous.
RUC. Repositorio da Universidade da Coruña
Publicación Contribución de congreso (ConferenceOutput)

Mapping the Unknown: A New Approach to Open-World Video Recognition

  • Parga, César D.
  • Pardo, Xosé Manuel
  • Regueiro, Carlos V.
Presentada no International Conference on Pattern Recognition, 2024. Parte da serie: Lecture Notes in Computer Science (LNCS,volume 15309). This version of the article has been accepted for publication, after peer review and is subject to Springer Nature’s AM terms of use, but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: https://doi.org/10.1007/978-3-031-78189-6_10, [Abstract]: Intelligent agents must strive to (re)map the constantly changing environment in which they operate, in order to remain adaptive and efficient. In open-world recognition (OWR) a system has to: detect new emerging categories, recognize new instances of known categories, and continually update knowledge based on the data streams it receives, mostly unannotated. In this work, we propose a hybrid method to deal with OWR that combines deep feature embeddings with dynamic ensemble methods for a continuous reshaping of boundaries in feature space. Our approach is flexible to update to patterns in the border of what is already known (concept-drift), detect and create models for new categories, recover from mistakes, and mitigate catastrophic forgetting, even in semi-supervised contexts. As an application use case, we have considered the problem of semi-supervised video face recognition, where the spatial-temporal coherence is harnessed to augment data. Our experiments shown that the system responds adequately to the unknowns, adding models for new identities, and improving its performance., This work has received financial support from the Spanish government (project PID2020-119367RB-I00); from the Xunta de Galicia, Consellaría de Cultura, Educación e Ordenación Universitaria (2019–2022 ED431G-2019/04 and ED431G 2019/01, and competitive groups 2021–2024 ED431C 2021/48 and ED431C 2021/30), and from the European Regional Development Fund (ERDF/FEDER). César D. Parga has received financial support from the Xunta de Galicia and the European Union (European Social Fund - ESF).
RUC. Repositorio da Universidade da Coruña
Publicación Artículo científico (JournalArticle)

Ensemble and continual federated learning for classification tasks

  • Casado, Fernando E.
  • Lema, Dylan
  • Iglesias, Roberto
  • Regueiro, Carlos V.
  • Barro, Senén
Financiado para publicación en acceso aberto: CRUE-CSIC/Springer Nature., The raw data used in the experiments described in Sect. 7 and Appendix A can be found on the following URL: https://citius.usc.es/t/30., [Abstract]: Federated learning is the state-of-the-art paradigm for training a learning model collaboratively across multiple distributed devices while ensuring data privacy. Under this framework, different algorithms have been developed in recent years and have been successfully applied to real use cases. The vast majority of work in federated learning assumes static datasets and relies on the use of deep neural networks. However, in real-world problems, it is common to have a continual data stream, which may be non-stationary, leading to phenomena such as concept drift. Besides, there are many multi-device applications where other, non-deep strategies are more suitable, due to their simplicity, explainability, or generalizability, among other reasons. In this paper we present Ensemble and Continual Federated Learning, a federated architecture based on ensemble techniques for solving continual classification tasks. We propose the global federated model to be an ensemble, consisting of several independent learners, which are locally trained. Thus, we enable a flexible aggregation of heterogeneous client models, which may differ in size, structure, or even algorithmic family. This ensemble-based approach, together with drift detection and adaptation mechanisms, also allows for continual adaptation in situations where data distribution changes over time. In order to test our proposal and illustrate how it works, we have evaluated it in different tasks related to human activity recognition using smartphones., Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. This research has received financial support from AEI/FEDER (European Union) Grant Number PID2020-119367RB-I00, as well as the Consellería de Cultura, Educación e Universitade of Galicia (accreditation ED431G-2019/04, ED431G2019/01, and ED431C2018/29), and the European Regional Development Fund (ERDF). It has also been supported by the Ministerio de Universidades of Spain in the FPU 2017 program (FPU17/04154)., Xunta de Galicia; ED431G-2019/04, Xunta de Galicia; ED431G2019/01, Xunta de Galicia; ED431C2018/29
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publicación Artículo científico (JournalArticle)

Adaptive deformation control for elastic linear objects

  • Aghajanzadeh, O.
  • Aranda, M.
  • Corrales Ramon, J.A.
  • Cariou, C.
  • Lenain, R.
  • Mezouar, Y.
This paper addresses the general problem of deformable linear object manipulation. The main application we consider is in the field of agriculture, for plant grasping, but may have interests in other tasks such as human daily activities and industrial production. We specifically consider an elastic linear object where one of its endpoints is fixed, and another point can be grasped by a robotic arm. To deal with the mentioned problem, we propose a model-free method to control the state of an arbitrary point that can be at any place along the object’s length. Our approach allows the robot to manipulate the object without knowing any model parameters or offline information of the object’s deformation. An adaptive control strategy is proposed for regulating the state of any point automatically deforming the object into the desired location. A control law is developed to regulate the object’s shape thanks to the adaptive estimation of the system parameters and its states. This method can track a desired manipulation trajectory to reach the target point, which leads to a smooth deformation without drastic changes. A Lyapunov-based argument is presented for the asymptotic convergence of the system that shows the process’s stability and convergence to desired state values. To validate the controller, numerical simulations involving two different deformation models are conducted, and performances of the proposed algorithm are investigated through full-scale experiments. Copyright © 2022 Aghajanzadeh, Aranda, Corrales Ramon, Cariou, Lenain and Mezouar.