Hello,

MY NAME IS MAXIMILIEN SERVAJEAN

I am currently assistant professor at LIRMM, Université Paul Valéry Montpellier. I conduct my research within the ADVANSE team. I have been previously involed at British Telecom in modeling host behaviors over the network in a scalable approach using SPARK and, in particular, focusing on deviant behaviors.

I conducted my Ph.D. thesis under the supervision of Pr. Esther Pacitti and defended it in December 2014 at LIRMM (i.e. Laboratoire d'Informatique, Robotique et Micro-électronique de Montpellier) in the Zenith Team led by Patrick Valduriez.
I was then involved in a crowdsourcing project under the supervision of Alexis Joly. The context of this work is associated to the Pl@ntNet project and is about plants identification and is part of a collaboration with Dennis Shasha from NYU (author of more than 400 papers in data science).

My main research interests are:

Inference // Bayesian networks// Machine learning // Crowdsourcing // Information Retrieval // Diversification // Recommendation // Clouds, decentralized and multi-sites systems // Gamification

Publications

For more recent papers, check: dblp

ARTICLES IN INTERNATIONAL REFEREED JOURNAL

M. Servajean, R. Chailan, A. Joly. "Non-parametric Bayesian annotator combination." Information Science, 2018. //

M. Servajean, A. Joly, D. Shasha, J. Champ and E. Pacitti. "Crowdsourcing Thousands of Specialized Labels: a Bayesian active training approach" IEEE Transactions on Multimedia, 2017. //

M. Servajean, R. Akbarinia, E. Pacitti and S. Amer-Yahia. "Profile Diversity for Query Processing using User Recommendations" Information Systems, Elsevier, Volumes 48, Pages 44-63, 2015. //

M. Servajean, E. Pacitti, M. Liroz-Gistau, S. Amer-Yahia and A. El Abbadi. "Increasing Coverage in Distributed Search and Recommendation with Profile Diversity" Transactions on Large-Scale Data- and Knowledge-Centered Systems, LNCS, Volume 9430, Pages 115-144, 2015. //

ARTICLES IN REFEREED NATIONAL CONFERENCES OR WORKSHOP

M. Servajean, E. Pacitti, S. Amer-Yahia and P. Neveu. "Profile Diversity for Phenotyping Data Search and Recommendation" BDA, in Nantes, France, 2013. //

M. Servajean, E. Pacitti, M. Liroz Gistau, S. Amer-Yahia and A. El Abaddi. "Exploiting Diversification in Distributed Recommendation" BDA, in Grenoble, France, 2014. //

M. Servajean, E. Pacitti, M. Liroz-Gistau, A. Joly and J. Champ. "PlantRT: Multi-Site Diversified Search and Recommendation for Citizen Sciences" BDA, in Grenoble, France, 2014. //

hidden

ARTICLES IN REFEREED INTERNATIONAL CONFERENCES OR WORKSHOPS

B. Deneu, M. Servajean, C. Botella, A. Joly. "Location-based Species Recommendation using Co-occurrences and Environment- GeoLifeCLEF 2018 Challenge" CLEF (working notes), in Avignon, France, 2018. //

W. Ragheb, B. Moulahi, J. Azé, S. Bringay, M. Servajean. "Temporal Mood Variation: at the CLEF eRisk-2018 Tasks for Early Risk Detection on The Internet" CLEF (working notes), in Avignon, France, 2018. //

M. R. Bouadjenek, E. Pacitti, M. Servajean, F. Masseglia, A. El Abbadi. "A Distributed Collaborative Filtering Algorithm Using Multiple Data Sources." DBKDA, in Nice, France, 2018. //

M. Servajean, A. Joly, D. Shasha, J. Champ and E. Pacitti. "ThePlantGame: Actively Training Human Annotators for Domain-specific Crowdsourcing" ACM Multimedia, in Amsterdam, The Netherlands, 2016. //

J. Champ, T. Lorieul, M. Servajean and A. Joly.
"A Comparative Study of Fine-grained Classification Methods in the Context of the LifeCLEF Plant Identification Challenge 2015" CLEF (Working Notes), in Toulouse, France, 2015. //

M. Servajean, E. Pacitti, S. Amer-Yahia and P. Neveu. "Profile Diversity in Search and Recommendation" WWW Companion, in Rio de Janeiro, Brazil, 2013. //

M. Servajean, E. Pacitti, M. Liroz-Gistau, S. Amer-Yahia and A. El Abbadi. "Exploiting Diversification in Gossip-Based Recommendation" Globe, in München, Deutschland, 2014. //

Teaching

apprentissage bayes big data graphe java

PhD. Students

Waleed Ragheb

The topic of this work is about suicide detection. He is particularly focused in making machine learning and deep learning especially explainable for human understanding.

Benjamin Deneu

Interpretability of distribution models of plant species communities learned through deep learning - application to crop weeds in the context of agro-ecology.

Internships

Anthony Pigeat

This 3 month internship consisted in the study of NoSQL databases and more precisely of the MongoDB document oriented database and the Neo4J graph database. This internship led to the realization of a tutorial as well as a set of exercises to facilitate the use of these databases.

Rider Carrion Cleger

This six-month internship led to the creation of a fully generic multi-site document-sharing platform. Each site allows the creation and identification of users. A multi-site search engine lets you retrieve relevant documents regardless of where they are stored.

Yonatan Carranza Alarcon

This 6 month internship consisted in creating a mathematical model to randomly generate crowdsourcing data based on predefined user profiles. The realization of this model is based on the idea of distributions of a priori probability and random draws. Finally, the responses of each user are themselves drawn randomly according to the profile of each user.

Ghofrane Ben Hamed

This 6 month internship consisted in developing an Android version of The Plant Game, thus leading to more attendance. This mobile app relies on web services to access The Plant Game Services and publish results obtained locally.

Alban Lorillard & Raphael Barraud

This 3 month internship at BT, Adastral Park, Ipswich, consisted in developping a dynamic workflow management system in order to help building machine learning pipelines for anomaly detection.

Benjamin Deneu

This 6 month internship at LIRMM consists in proposing various machine learning models in order to predict species spatial distribution.

Mathilde Negri

A 6 month internship to analyse the ability of deep species distribution models to predict in novel geographical area.

Lucas Bernigaud Samatan

A 3 month internship to study the variation in species predictions near and within agricultural plots in order to better understand the interactions between farming practice and biodiversity.

Production

The Plant Game

More than 60,000 users (Google Analytics). It enables to label observations generated by the Pl@ntnet platform (15,000 users daily from more than 170 countries and 3 million total download). It relies on various techniques such as Bayesian networks and deep convolutional neural networks.

Check here.

POSTERS

A Personalized and Diversified Model for Decentralized Scientific Search and Recommendation

Download here.

Multi-Site Diversified Search and Recommendation

Download here.

PROTOTYPES

Document Recommendation Tool

Check here.

Plant Recommendation Tool

Check here.

Contact me

TO maximiliense_at_gmail.com

ADDRESS Bâtiment 5, LIRMM, 860 Rue de St Priest 34095 Montpellier France