-
Linearized Implicit Methods Based on a Single-Layer Neural Network: Application to Keller-Segel Models
Authors:
M. Benzakour Amine
Abstract:
This paper is concerned with numerical approximation of some two-dimensional Keller-Segel chemotaxis models, especially those generating pattern formations. The numerical resolution of such nonlinear parabolic-parabolic or parabolic-elliptic systems of partial differential equations consumes a significant computational time when solved with fully implicit schemes. Standard linearized semi-implicit…
▽ More
This paper is concerned with numerical approximation of some two-dimensional Keller-Segel chemotaxis models, especially those generating pattern formations. The numerical resolution of such nonlinear parabolic-parabolic or parabolic-elliptic systems of partial differential equations consumes a significant computational time when solved with fully implicit schemes. Standard linearized semi-implicit schemes, however, require reasonable computational time, but suffer from lack of accuracy. In this work, two methods based on a single-layer neural network are developed to build linearized implicit schemes: a basic one called the each step training linearized implicit (ESTLI) method and a more efficient one, the selected steps training linearized implicit (SSTLI) method. The proposed schemes, which make use also of a spatial finite volume method with a hybrid difference scheme approximation for convection-diffusion fluxes, are first derived for a chemotaxis system arising in embryology. The convergence of the numerical solutions to a corresponding weak solution of the studied system is established. Then the proposed methods are applied to a number of chemotaxis models, and several numerical tests are performed to illustrate their accuracy, efficiency and robustness. Generalization of the developed methods to other nonlinear partial differential equations is straightforward.
△ Less
Submitted 27 October, 2020; v1 submitted 7 April, 2020;
originally announced April 2020.
-
A corrected decoupled scheme for chemotaxis models
Authors:
M. Akhmouch,
M. Benzakour Amine
Abstract:
The main purpose of this paper is to present a new corrected decoupled scheme combined with a spatial finite volume method for chemotaxis models. First, we derive the scheme for a parabolic-elliptic chemotaxis model arising in embryology. We then establish the existence and uniqueness of the numerical solution, and we prove that it converges to a corresponding weak solution for the studied model.…
▽ More
The main purpose of this paper is to present a new corrected decoupled scheme combined with a spatial finite volume method for chemotaxis models. First, we derive the scheme for a parabolic-elliptic chemotaxis model arising in embryology. We then establish the existence and uniqueness of the numerical solution, and we prove that it converges to a corresponding weak solution for the studied model. In the last section, several numerical tests are presented by applying our approach to a number of chemotaxis systems. The obtained numerical results demonstrate the efficiency of the proposed scheme and its effectiveness to capture different forms of spatial patterns.
△ Less
Submitted 26 February, 2020;
originally announced February 2020.
-
SoccerNet: A Scalable Dataset for Action Spotting in Soccer Videos
Authors:
Silvio Giancola,
Mohieddine Amine,
Tarek Dghaily,
Bernard Ghanem
Abstract:
In this paper, we introduce SoccerNet, a benchmark for action spotting in soccer videos. The dataset is composed of 500 complete soccer games from six main European leagues, covering three seasons from 2014 to 2017 and a total duration of 764 hours. A total of 6,637 temporal annotations are automatically parsed from online match reports at a one minute resolution for three main classes of events (…
▽ More
In this paper, we introduce SoccerNet, a benchmark for action spotting in soccer videos. The dataset is composed of 500 complete soccer games from six main European leagues, covering three seasons from 2014 to 2017 and a total duration of 764 hours. A total of 6,637 temporal annotations are automatically parsed from online match reports at a one minute resolution for three main classes of events (Goal, Yellow/Red Card, and Substitution). As such, the dataset is easily scalable. These annotations are manually refined to a one second resolution by anchoring them at a single timestamp following well-defined soccer rules. With an average of one event every 6.9 minutes, this dataset focuses on the problem of localizing very sparse events within long videos. We define the task of spotting as finding the anchors of soccer events in a video. Making use of recent developments in the realm of generic action recognition and detection in video, we provide strong baselines for detecting soccer events. We show that our best model for classifying temporal segments of length one minute reaches a mean Average Precision (mAP) of 67.8%. For the spotting task, our baseline reaches an Average-mAP of 49.7% for tolerances $δ$ ranging from 5 to 60 seconds. Our dataset and models are available at https://silviogiancola.github.io/SoccerNet.
△ Less
Submitted 22 April, 2018; v1 submitted 12 April, 2018;
originally announced April 2018.
-
Vers une interface pour l enrichissement des requetes en arabe dans un systeme de recherche d information
Authors:
Abderrahim Mohammed El Amine
Abstract:
This presentation focuses on the automatic expansion of Arabic request using morphological analyzer and Arabic Wordnet. The expanded request is sent to Google.
This presentation focuses on the automatic expansion of Arabic request using morphological analyzer and Arabic Wordnet. The expanded request is sent to Google.
△ Less
Submitted 6 February, 2014;
originally announced February 2014.
-
The Number of Terms and Documents for Pseudo-Relevant Feedback for Ad-hoc Information Retrieval
Authors:
Abderrahim Mohammed El Amine,
Benameur Said,
Abderrahim Mohammed Alaeddine
Abstract:
In Information Retrieval System (IRS), the Automatic Relevance Feedback (ARF) is a query reformulation technique that modifies the initial one without the user intervention. It is applied mainly through the addition of terms coming from the external resources such as the ontologies and or the results of the current research. In this context we are mainly interested in the local analysis technique…
▽ More
In Information Retrieval System (IRS), the Automatic Relevance Feedback (ARF) is a query reformulation technique that modifies the initial one without the user intervention. It is applied mainly through the addition of terms coming from the external resources such as the ontologies and or the results of the current research. In this context we are mainly interested in the local analysis technique for the ARF in ad-hoc IRS on Arabic documents. In this article, we have examined the impact of the variation of the two parameters implied in this technique, that is to say, the number of the documents «D» and the number of terms «T», on an Arabic IRS performance. The experimentation, carried out on an Arabic corpus text, enables us to deduce that there are queries which are not easily improvable with the query reformulation. In addition, the success of the ARF is due mainly to the selection of a sufficient number of documents D and to the extraction of a very reduced set of relevant terms T for retrieval.
△ Less
Submitted 17 June, 2013;
originally announced June 2013.
