Skip to main content

APRENDIZAJE MAQUINA "GLOCAL" Y CONTINUO PARA UNA SOCIEDAD DE DISPOSITIVOS INTELIGENTES

TIN2017-90135-R

•
Funding agency name Agencia Estatal de Investigación
Funding agency acronym AEI
Program Programa Estatal de I+D+i Orientado a los Retos de la Sociedad
Call Proyectos I+D+i Retos de Investigación
Call year 2017
State Plan 2017-2020 Plan Estatal de Investigación Científica y Técnica y de Innovación
Beneficiary centre UNIVERSIDADE DE SANTIAGO DE COMPOSTELA

Resultados relacionados

Resultados totales (Incluyendo duplicados): 10
Encontrada(s) 1 página(s)

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

Robust Step Counting for Inertial Navigation with Mobile Phones

  • Rodríguez García, Germán
  • Estévez Casado, Fernando
  • Vázquez Regueiro, Carlos
  • Nieto, Adrián
Mobile phones are increasingly used for purposes that have nothing to do with phone calls or simple data transfers, and one such use is indoor inertial navigation. Nevertheless, the development of a standalone application able to detect the displacement of the user starting only from the data provided by the most common inertial sensors in the mobile phones (accelerometer, gyroscope and magnetometer), is a complex task. This complexity lies in the hardware disparity, noise on data, and mostly the many movements that the mobile phone can experience and which have nothing to do with the physical displacement of the owner. In our case, we describe a proposal, which, after using quaternions and a Kalman filter to project the sensors readings into an Earth Centered inertial reference system, combines a classic Peak-valley detector with an ensemble of SVMs (Support Vector Machines) and a standard deviation based classifier. Our proposal is able to identify and filter out those segments of signal that do not correspond to the behavior of “walking”, and thus achieve a robust detection of the physical displacement and counting of steps. We have performed an extensive experimental validation of our proposal using a dataset with 140 records obtained from 75 different people who were not connected to this research
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publication Artículo científico (JournalArticle)

Concept drift detection and adaptation for federated and continual learning

  • Estévez Casado, Fernando
  • Lema Pais, Dylan
  • Fernández Criado, Marcos
  • Iglesias Rodríguez, Roberto
  • Vázquez Regueiro, Carlos
  • Barro Ameneiro, Senén
Smart devices, such as smartphones, wearables, robots, and others, can collect vast amounts of data from their environment. This data is suitable for training machine learning models, which can significantly improve their behavior, and therefore, the user experience. Federated learning is a young and popular framework that allows multiple distributed devices to train deep learning models collaboratively while preserving data privacy. Nevertheless, this approach may not be optimal for scenarios where data distribution is non-identical among the participants or changes over time, causing what is known as concept drift. Little research has yet been done in this field, but this kind of situation is quite frequent in real life and poses new challenges to both continual and federated learning. Therefore, in this work, we present a new method, called Concept-Drift-Aware Federated Averaging (CDA-FedAvg). Our proposal is an extension of the most popular federated algorithm, Federated Averaging (FedAvg), enhancing it for continual adaptation under concept drift. We empirically demonstrate the weaknesses of regular FedAvg and prove that CDA-FedAvg outperforms it in this type of scenario
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publication Artículo científico (JournalArticle)

Robust and Fast Scene Recognition in Robotics Through the Automatic Identification of Meaningful Images

  • Santos Saavedra, David
  • López López, Eric
  • Pardo López, Xosé Manuel
  • Iglesias Rodríguez, Roberto
  • Barro Ameneiro, Senén
  • Fernández Vidal, Xosé Ramón
Scene recognition is still a very important topic in many fields, and that is definitely the case in robotics. Nevertheless, this task is view-dependent, which implies the existence of preferable directions when recognizing a particular scene. Both in human and computer vision-based classification, this actually often turns out to be biased. In our case, instead of trying to improve the generalization capability for different view directions, we have opted for the development of a system capable of filtering out noisy or meaningless images while, on the contrary, retaining those views from which is likely feasible that the correct identification of the scene can be made. Our proposal works with a heuristic metric based on the detection of key points in 3D meshes (Harris 3D). This metric is later used to build a model that combines a Minimum Spanning Tree and a Support Vector Machine (SVM). We have performed an extensive number of experiments through which we have addressed (a) the search for efficient visual descriptors, (b) the analysis of the extent to which our heuristic metric resembles the human criteria for relevance and, finally, (c) the experimental validation of our complete proposal. In the experiments, we have used both a public image database and images collected at our research center
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publication Artículo científico (JournalArticle)

Artificial intelligence within the interplay between natural and artificial computation: Advances in data science, trends and applications

  • Górriz, Juan M.
  • Ramírez, Javier
  • Ortíz, Andrés
  • Martínez Murcia, Francisco J.
  • Segovia, Fermin
  • Suckling, John
  • Leming, Matthew
  • Zhang, Yu-Dong
  • Álvarez Sánchez, Ramón
  • Bologna, Guido
  • Bonomini, Paula
  • Estévez Casado, Fernando
  • Charte, David
  • Charte, Francisco
  • Contreras, Ricardo
  • Cuesta Infante, Alfredo
  • Duro, Richard J.
  • Fernández Caballero, Antonio
  • Fernández Jover, Eduardo
  • Gómez Vilda, Pedro
  • Graña, Manuel
  • Herrera, Francisco
  • Iglesias Rodríguez, Roberto
  • Lekova, Anna
  • Lope, Javier de
  • López Rubio, Ezequiel
  • Martínez Tomás, Rafael
  • Molina Cabello, Miguel A.
  • Montemayor, Antonio S.
  • Novais, Paulo
  • Palacios Alonso, Daniel
  • Pantrigo, Juan J.
  • Payne, Bryson R.
  • Paz López, Félix de la
  • Pinninghoff, María Angélica
  • Rincón, Mariano
  • Santos, José
  • Thurnhofer Hemsi, Karl
  • Tsanas, Athanasios
  • Varela, Ramiro
  • Ferrández, Jose M.
Artificial intelligence and all its supporting tools, e.g. machine and deep learning in computational intelligence-based systems, are rebuilding our society (economy, education, life-style, etc.) and promising a new era for the social welfare state. In this paper we summarize recent advances in data science and artificial intelligence within the interplay between natural and artificial computation. A review of recent works published in the latter field and the state the art are summarized in a comprehensive and self-contained way to provide a baseline framework for the international community in artificial intelligence. Moreover, this paper aims to provide a complete analysis and some relevant discussions of the current trends and insights within several theoretical and application fields covered in the essay, from theoretical models in artificial intelligence and machine learning to the most prospective applications in robotics, neuroscience, brain computer interfaces, medicine and society, in general
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publication Artículo científico (JournalArticle)

Walking Recognition in Mobile Devices

  • Estévez Casado, Fernando
  • Rodríguez García, Germán
  • Iglesias Rodríguez, Roberto
  • Vázquez Regueiro, Carlos
  • Barro Ameneiro, Senén
  • Canedo Rodríguez, Adrián
Presently, smartphones are used more and more for purposes that have nothing to do with phone calls or simple data transfers. One example is the recognition of human activity, which is relevant information for many applications in the domains of medical diagnosis, elderly assistance, indoor localization, and navigation. The information captured by the inertial sensors of the phone (accelerometer, gyroscope, and magnetometer) can be analyzed to determine the activity performed by the person who is carrying the device, in particular in the activity of walking. Nevertheless, the development of a standalone application able to detect the walking activity starting only from the data provided by these inertial sensors is a complex task. This complexity lies in the hardware disparity, noise on data, and mostly the many movements that the smartphone can experience and which have nothing to do with the physical displacement of the owner. In this work, we explore and compare several approaches for identifying the walking activity. We categorize them into two main groups: the first one uses features extracted from the inertial data, whereas the second one analyzes the characteristic shape of the time series made up of the sensors readings. Due to the lack of public datasets of inertial data from smartphones for the recognition of human activity under no constraints, we collected data from 77 different people who were not connected to this research. Using this dataset, which we published online, we performed an extensive experimental validation and comparison of our proposals
RUC. Repositorio da Universidade da Coruña
Publication Artículo científico (JournalArticle)

Robust step counting for inertial navigation with mobile phones

  • Rodríguez García, Germán
  • Casado, Fernando E.
  • Iglesias Rodríguez, Roberto
  • Regueiro, Carlos V.
  • Nieto, Adrián
[Abstract]: Mobile phones are increasingly used for purposes that have nothing to do with phone calls or simple data transfers, and one such use is indoor inertial navigation. Nevertheless, the development of a standalone application able to detect the displacement of the user starting only from the data provided by the most common inertial sensors in the mobile phones (accelerometer, gyroscope and magnetometer), is a complex task. This complexity lies in the hardware disparity, noise on data, and mostly the many movements that the mobile phone can experience and which have nothing to do with the physical displacement of the owner. In our case, we describe a proposal, which, after using quaternions and a Kalman filter to project the sensors readings into an Earth Centered inertial reference system, combines a classic Peak-valley detector with an ensemble of SVMs (Support Vector Machines) and a standard deviation based classifier. Our proposal is able to identify and filter out those segments of signal that do not correspond to the behavior of “walking”, and thus achieve a robust detection of the physical displacement and counting of steps. We have performed an extensive experimental validation of our proposal using a dataset with 140 records obtained from 75 different people who were not connected to this research., This research was funded by (AEI/FEDER, UE) grant number TIN2017-90135-R, as well as the Consellería de Cultura, Educación e Ordenación Universitaria (accreditation 2016–2019, ED431G/01 and ED431G/08 and reference competitive group 2014–2017, GRC2014/030), the European Regional Development Fund (ERDF) and was supported by the Ministerio de Economía, Industria y Competitividad in the program Industrial PhD 2014., Xunta de Galicia; ED431G/01, Xunta de Galicia; ED431G/08, Xunta de Galicia; GRC2014/030
RUC. Repositorio da Universidade da Coruña
Publication Artículo científico (JournalArticle)

Concept Drift Detection and Adaptation for Federated and Continual Learning

  • Casado, Fernando E.
  • Lema, Dylan
  • Criado, Marcos F.
  • Iglesias, Roberto
  • Regueiro, Carlos V.
  • Barro, Senén
[Abstract] Smart devices, such as smartphones, wearables, robots, and others, can collect vast amounts of data from their environment. This data is suitable for training machine learning models, which can significantly improve their behavior, and therefore, the user experience. Federated learning is a young and popular framework that allows multiple distributed devices to train deep learning models collaboratively while preserving data privacy. Nevertheless, this approach may not be optimal for scenarios where data distribution is non-identical among the participants or changes over time, causing what is known as concept drift. Little research has yet been done in this field, but this kind of situation is quite frequent in real life and poses new challenges to both continual and federated learning. Therefore, in this work, we present a new method, called Concept-Drift-Aware Federated Averaging (CDA-FedAvg). Our proposal is an extension of the most popular federated algorithm, Federated Averaging (FedAvg), enhancing it for continual adaptation under concept drift. We empirically demonstrate the weaknesses of regular FedAvg and prove that CDA-FedAvg outperforms it in this type of scenario., Xunta de Galicia; ED431G/01, Xunta de Galicia; ED431G/08, Xunta de Galicia; ED431C2018/29, Xunta de Galicia; ED431F2018/02, Esta investigación ha recibido apoyo financiero de la AEI/FEDER (UE) con número de subvención TIN2017-90135-R, así como de la Consellería de Cultura, Educación e Ordenación Universitaria de Galicia (acreditación 2016-2019, ED431G/01 y ED431G/08, grupo competitivo de referencia ED431C2018/29, y subvención ED431F2018/02), y del Fondo Europeo de Desarrollo Regional (FEDER). También ha sido apoyado por el Ministerio de Universidades de España en el programa FPU 2017 (FPU17/04154).
RUC. Repositorio da Universidade da Coruña
Publication Artículo científico (JournalArticle)

Walking Recognition in Mobile Devices

  • Casado, Fernando E.
  • Rodríguez García, Germán
  • Iglesias Rodríguez, Roberto
  • Regueiro, Carlos V.
  • Barro, Senén
  • Canedo-Rodríguez, Adrián
[Abstract] Presently, smartphones are used more and more for purposes that have nothing to do with phone calls or simple data transfers. One example is the recognition of human activity, which is relevant information for many applications in the domains of medical diagnosis, elderly assistance, indoor localization, and navigation. The information captured by the inertial sensors of the phone (accelerometer, gyroscope, and magnetometer) can be analyzed to determine the activity performed by the person who is carrying the device, in particular in the activity of walking. Nevertheless, the development of a standalone application able to detect the walking activity starting only from the data provided by these inertial sensors is a complex task. This complexity lies in the hardware disparity, noise on data, and mostly the many movements that the smartphone can experience and which have nothing to do with the physical displacement of the owner. In this work, we explore and compare several approaches for identifying the walking activity. We categorize them into two main groups: the first one uses features extracted from the inertial data, whereas the second one analyzes the characteristic shape of the time series made up of the sensors readings. Due to the lack of public datasets of inertial data from smartphones for the recognition of human activity under no constraints, we collected data from 77 different people who were not connected to this research. Using this dataset, which we published online, we performed an extensive experimental validation and comparison of our proposals., This research has received financial support from AEI/FEDER (European Union) grant number TIN2017-90135-R, as well as the Consellería de Cultura, Educación e Ordenación Universitaria of Galicia (accreditation 2016–2019, ED431G/01 and ED431G/08, reference competitive group ED431C2018/29, and grant ED431F2018/02), and the European Regional Development Fund (ERDF). It has also been supported by the Ministerio de Educación, Cultura y Deporte of Spain in the FPU 2017 program (FPU17/04154), and the Ministerio de Economía, Industria y Competitividad in the Industrial PhD 2014 program (DI-14-06920), Xunta de Galicia; ED431G/01, Xunta de Galicia; ED431G/08, Xunta de Galicia; ED431C2018/29, Xunta de Galicia; ED431F2018/02
RUC. Repositorio da Universidade da Coruña
Publication Tesis doctoral (DoctoralThesis)

Incremental Learning Through Unsupervised Adaptation in Video Face Recognition

  • López-López, Eric
[Resumo] Durante a última década, os métodos baseados en deep learning trouxeron un salto significativo no rendemento dos sistemas de visión artificial. Unha das claves neste éxito foi a creación de grandes conxuntos de datos perfectamente etiquetados para usar durante o adestramento. En certa forma, as redes de deep learning resumen esta enorme cantidade datos en prácticos vectores multidimensionais. Por este motivo, cando as diferenzas entre os datos de adestramento e os adquiridos durante o funcionamento dos sistemas (debido a factores como o contexto de adquisición) son especialmente notorias, as redes de deep learning son susceptibles de sufrir degradación no rendemento. Mentres que a solución inmediata a este tipo de problemas sería a de recorrer a unha recolección adicional de imaxes, co seu correspondente proceso de etiquetado, esta dista moito de ser óptima. A gran cantidade de posibles variacións que presenta o mundo visual converten rápido este enfoque nunha tarefa sen fin. Máis aínda cando existen aplicacións específicas nas que esta acción é difícil, ou incluso imposible, de realizar debido a problemas de custos ou de privacidade. Esta tese propón abordar todos estes problemas usando a perspectiva da adaptación. Así, a hipótese central consiste en asumir que é posible utilizar os datos non etiquetados adquiridos durante o funcionamento para mellorar o rendemento que obteríamos con sistemas de recoñecemento xerais. Para isto, e como proba de concepto, o campo de estudo da tese restrinxiuse ao recoñecemento de caras. Esta é unha aplicación paradigmática na cal o contexto de adquisición pode ser especialmente relevante. Este traballo comeza examinando as diferenzas intrínsecas entre algúns dos contextos específicos nos que se pode necesitar o recoñecemento de caras e como estas afectan ao rendemento. Desta maneira, comparamos distintas bases de datos (xunto cos seus contextos) entre elas, usando algúns dos descritores de características máis avanzados e así determinar a necesidade real de adaptación. A partir desta punto, pasamos a presentar o método novo, que representa a principal contribución da tese: o Dynamic Ensemble of SVM (De-SVM). Este método implementa a capacidade de adaptación utilizando unha aprendizaxe incremental non supervisada na que as súas propias predicións se usan como pseudo-etiquetas durante as actualizacións (a estratexia de auto-adestramento). Os experimentos realizáronse baixo condicións de vídeo-vixilancia, un exemplo paradigmático dun contexto moi específico no que os procesos de etiquetado son particularmente complicados. As ideas claves de De-SVM probáronse en diferentes sub-problemas de recoñecemento de caras: a verificación de caras e recoñecemento de caras en conxunto pechado e en conxunto aberto. Os resultados acadados mostran un comportamento prometedor en termos de adquisición de coñecemento sen supervisión así como robustez contra impostores. Ademais, este rendemento é capaz de superar a outros métodos do estado da arte que non posúen esta capacidade de adaptación., [Resumen] Durante la última década, los métodos basados en deep learning trajeron un salto significativo en el rendimiento de los sistemas de visión artificial. Una de las claves en este éxito fue la creación de grandes conjuntos de datos perfectamente etiquetados para usar durante el entrenamiento. En cierta forma, las redes de deep learning resumen esta enorme cantidad datos en prácticos vectores multidimensionales. Por este motivo, cuando las diferencias entre los datos de entrenamiento y los adquiridos durante el funcionamiento de los sistemas (debido a factores como el contexto de adquisición) son especialmente notorias, las redes de deep learning son susceptibles de sufrir degradación en el rendimiento. Mientras que la solución a este tipo de problemas es recurrir a una recolección adicional de imágenes, con su correspondiente proceso de etiquetado, esta dista mucho de ser óptima. La gran cantidad de posibles variaciones que presenta el mundo visual convierten rápido este enfoque en una tarea sin fin. Más aún cuando existen aplicaciones específicas en las que esta acción es difícil, o incluso imposible, de realizar; debido a problemas de costes o de privacidad. Esta tesis propone abordar todos estos problemas usando la perspectiva de la adaptación. Así, la hipótesis central consiste en asumir que es posible utilizar los datos no etiquetados adquiridos durante el funcionamiento para mejorar el rendimiento que se obtendría con sistemas de reconocimiento generales. Para esto, y como prueba de concepto, el campo de estudio de la tesis se restringió al reconocimiento de caras. Esta es una aplicación paradigmática en la cual el contexto de adquisición puede ser especialmente relevante. Este trabajo comienza examinando las diferencias entre algunos de los contextos específicos en los que se puede necesitar el reconocimiento de caras y así como sus efectos en términos de rendimiento. De esta manera, comparamos distintas ba ses de datos (y sus contextos) entre ellas, usando algunos de los descriptores de características más avanzados para así determinar la necesidad real de adaptación. A partir de este punto, pasamos a presentar el nuevo método, que representa la principal contribución de la tesis: el Dynamic Ensemble of SVM (De- SVM). Este método implementa la capacidad de adaptación utilizando un aprendizaje incremental no supervisado en la que sus propias predicciones se usan cómo pseudo-etiquetas durante las actualizaciones (la estrategia de auto-entrenamiento). Los experimentos se realizaron bajo condiciones de vídeo-vigilancia, un ejemplo paradigmático de contexto muy específico en el que los procesos de etiquetado son particularmente complicados. Las ideas claves de De- SVM se probaron en varios sub-problemas del reconocimiento de caras: la verificación de caras y reconocimiento de caras de conjunto cerrado y conjunto abierto. Los resultados muestran un comportamiento prometedor en términos de adquisición de conocimiento así como de robustez contra impostores. Además, este rendimiento es capaz de superar a otros métodos del estado del arte que no poseen esta capacidad de adaptación., [Abstract] In the last decade, deep learning has brought an unprecedented leap forward for computer vision general classification problems. One of the keys to this success is the availability of extensive and wealthy annotated datasets to use as training samples. In some sense, a deep learning network summarises this enormous amount of data into handy vector representations. For this reason, when the differences between training datasets and the data acquired during operation (due to factors such as the acquisition context) are highly marked, end-to-end deep learning methods are susceptible to suffer performance degradation. While the immediate solution to mitigate these problems is to resort to an additional data collection and its correspondent annotation procedure, this solution is far from optimal. The immeasurable possible variations of the visual world can convert the collection and annotation of data into an endless task. Even more when there are specific applications in which this additional action is difficult or simply not possible to perform due to, among other reasons, cost-related problems or privacy issues. This Thesis proposes to tackle all these problems from the adaptation point of view. Thus, the central hypothesis assumes that it is possible to use operational data with almost no supervision to improve the performance we would achieve with general-purpose recognition systems. To do so, and as a proof-of-concept, the field of study of this Thesis is restricted to face recognition, a paradigmatic application in which the context of acquisition can be especially relevant. This work begins by examining the intrinsic differences between some of the face recognition contexts and how they directly affect performance. To do it, we compare different datasets, and their contexts, against each other using some of the most advanced feature representations available to determine the actual need for adaptation. From this point, we move to present the novel method, representing the central contribution of the Thesis: the Dynamic Ensembles of SVM (De-SVM). This method implements the adaptation capabilities by performing unsupervised incremental learning using its own predictions as pseudo-labels for the update decision (the self-training strategy). Experiments are performed under video surveillance conditions, a paradigmatic example of a very specific context in which labelling processes are particularly complicated. The core ideas of De-SVM are tested in different face recognition sub-problems: face verification and, the more complex, general closed- and open-set face recognition. In terms of the achieved results, experiments have shown a promising behaviour in terms of both unsupervised knowledge acquisition and robustness against impostors, surpassing the performances achieved by state-of-the-art non-adaptive methods., Funding and Technical Resources For the successful development of this Thesis, it was necessary to rely on series of indispensable means included in the following list: • Working material, human and financial support primarily by the CITIC and the Computer Architecture Group of the University of A Coruña and CiTIUS of University of Santiago de Compostela, along with a PhD grant funded by Xunta the Galicia and the European Social Fund. • Access to bibliographical material through the library of the University of A Coruña. • Additional funding through the following research projects: State funding by the Ministry of Economy and Competitiveness of Spain (project TIN2017-90135-R MINECO, FEDER),
RUC. Repositorio da Universidade da Coruña
Publication Artículo científico (JournalArticle)

Towards a Self-Sufficient Face Verification System

  • López-López, Eric
  • Regueiro, Carlos V.
  • Pardo, Xosé Manuel
  • Franco, Annalisa
  • Lumini, Alessandra
Financiado para publicación en acceso aberto: Universidade da Coruña/CISUG, [Abstract] The absence of a previous collaborative manual enrolment represents a significant handicap towards designing a face verification system for face re-identification purposes. In this scenario, the system must learn the target identity incrementally, using data from the video stream during the operational authentication phase. So, manual labelling cannot be assumed apart from the first few frames. On the other hand, even the most advanced methods trained on large-scale and unconstrained datasets suffer performance degradation when no adaptation to specific contexts is performed. This work proposes an adaptive face verification system, for the continuous re-identification of target identity, within the framework of incremental unsupervised learning. Our Dynamic Ensemble of SVM is capable of incorporating non-labelled information to improve the performance of any model, even when its initial performance is modest. The proposal uses the self-training approach and is compared against other classification techniques within this same approach. Results show promising behaviour in terms of both knowledge acquisition and impostor robustness., This work has received financial support from the Spanish government (project TIN2017-90135-R MINECO (FEDER)), from The Consellaría de Cultura, Educación e Ordenación Universitaria (accreditations 2016–2019, EDG431G/01 and ED431G/08), and reference competitive groups (2017–2020, and ED431C 2017/04), and from the European Regional Development Fund (ERDF). Eric López-López has received financial support from the Xunta de Galicia and the European Union (European Social Fund – ESF), Xunta de Galicia; EDG431G/01, Xunta de Galicia; ED431G/08, Xunta de Galicia; ED431C 2017/04