Postdoc, Nancy.| Entreprise/Organisme : | Institut Elie Cartan de Lorraine | | Niveau d'études : | Doctorat | | Sujet : | The AIHD project (https://aihd.loria.fr) is an interdisciplinary project that aims
at providing data-driven Artificial Intelligence (AI) solutions for medical diagnosis and
prognosis applications related to heart diseases, and more specifically Atrial Fibrillation
(AF).
One application consists in the automated detection of AF episodes from a short
single lead electrocardiogram (ECG) recording obtained with a mobile device. This could
lead to the early detection of paroxysmal atrial fibrillation episodes, therefore allowing a
better care and preventing adverse events such as stroke or heart failure.
From the mathematical point of view, ECG recording are multi-dimensional time
series. We propose to develop and improve statistical learning methods using new features
developed for multivariate time series analysis and encoding dependencies between the
components, the iterated integral signature (IIS) [1].
The signature method has shown its effectiveness in a large range of applications : as
classification of hand written characters in different languages as arabic [4] or chinese [3],
clustering of financial time series [2].
The latest up-to-date methods using the IIS are neural networks based and inherit
the accuracy and computational efficiency of deep learning approaches.
In this post doc we intend to explore IIS for semi-supervised learning and clustering
and compare it to other classical approaches.
Références
[1] I. Chevyrev and A. Kormilitzin. A primer on the signature method in machine learning,
2016. http ://arxiv.org/abs/1603.03788 arXiv :1603.03788.
[2] L.J. Gurkyó, T. Lyons, M. Kontkowski, and J. Field. Extracting information from the signa-
ture of a financial data stream, 2014. http ://arxiv.org/abs/1307.7244 arXiv :1307.7244.
[3] S. Lai, L. Jin, and W. Yang. Online signature verification using recurrent neural network and
length-normalized path signature, 2017. http ://arxiv.org/abs/1705.06849 arXiv :1705.06849.
[4] D. Wilson-Nunn, T. Lyons, A. Papavasiliou, and H. Ni. A path signature approach to online
arabic handwriting recognition. In 2018 IEEE 2nd International Workshop on Arabic and
Derived Script Analysis and Recognition (ASAR), pages 135–139. IEEE, 2018. | | Date de début : | 01/04/2020 | | Durée du contrat : | 12 mois | | Rémunération : | 2200E | | Secteur d'activité : | Statistique/machine learning | | Description : | Context
The AIHD project (https://aihd.loria.fr) is an interdisciplinary project that aims
at providing data-driven Artificial Intelligence (AI) solutions for medical diagnosis and
prognosis applications related to heart diseases, and more specifically Atrial Fibrillation
(AF).
One application consists in the automated detection of AF episodes from a short
single lead electrocardiogram (ECG) recording obtained with a mobile device. This could
lead to the early detection of paroxysmal atrial fibrillation episodes, therefore allowing a
better care and preventing adverse events such as stroke or heart failure.
From the mathematical point of view, ECG recording are multi-dimensional time
series. We propose to develop and improve statistical learning methods using new features
developed for multivariate time series analysis and encoding dependencies between the
components, the iterated integral signature (IIS) [1].
The signature method has shown its effectiveness in a large range of applications : as
classification of hand written characters in different languages as arabic [4] or chinese [3],
clustering of financial time series [2].
The latest up-to-date methods using the IIS are neural networks based and inherit
the accuracy and computational efficiency of deep learning approaches.
In this post doc we intend to explore IIS for semi-supervised learning and clustering
and compare it to other classical approaches.
Références
[1] I. Chevyrev and A. Kormilitzin. A primer on the signature method in machine learning,
2016. http ://arxiv.org/abs/1603.03788 arXiv :1603.03788.
[2] L.J. Gurkyó, T. Lyons, M. Kontkowski, and J. Field. Extracting information from the signa-
ture of a financial data stream, 2014. http ://arxiv.org/abs/1307.7244 arXiv :1307.7244.
[3] S. Lai, L. Jin, and W. Yang. Online signature verification using recurrent neural network and
length-normalized path signature, 2017. http ://arxiv.org/abs/1705.06849 arXiv :1705.06849.
[4] D. Wilson-Nunn, T. Lyons, A. Papavasiliou, and H. Ni. A path signature approach to online
arabic handwriting recognition. In 2018 IEEE 2nd International Workshop on Arabic and
Derived Script Analysis and Recognition (ASAR), pages 135–139. IEEE, 2018.ç | | En savoir plus : | https://docs.google.com/viewer?a=v&pid=sites&srcid=ZGVmYXVsdGRvbWFpbnxtYXJpYW5uZWNsYXVzZWx8Z3g6MjY2Y
PostDocMirabelle.pdf | | Contact : | marianne.clausel@univ-lorraine.fr |
|
