University Pierre et Marie Curie - LIP6 - UPMC 4 Place Jussieu - 75005 PARIS Tel : +33 1 44 27 84 23 Fax : +33 1 44 27 70 00 ludovic[dot]denoyer[at]lip6[dot]fr
NOW:
- Research Scientist at Facebook Artificial Intelligence Research - Paris
Previously:
- Full Professor in the Machine Learning and Information Access Team
- Co-head of the Data-Science Master Data, Machine Learning and Knowledge (DAC)
- Chargé de Mission Open Data/Research auprès de la présidence de l'UPMC
- Habilitation to Drive Research: http://www-connex.lip6.fr/~denoyer/HDR.pdf
- CV (PDF) : cv_july_2017
Slides
- Présentation Criteo2017 -- Sequential Budgeted Learning with focus on Cost-Sensitive Features Acquisition and Learning Time-Efficient Deep Architectures -- slides_criteo
- Présentation Google 2017 -- Sequential Budgeted Learning with focus on Deep Sequential Neural Networks and Budgeted Option Discovery -- PDF file: google_2017
- Présentation Critéo 2016 -- Learning Graph Embeddings -- criteo_2017
- Séminaire Facebook FAIR 2016 -- Sequential Budgeted Learning with focus on Sequential Information Acquisition --facebook_2016
Softwares
- A RL package for PyTorch : https://github.com/ludc/rl
- A RL simple package for Torch and OpenAI Gym : https://github.com/ludc/rltorch
- Source code of the Social Network Embeddings paper : https://github.com/ludc/social_network_diffusion_embeddings
- A simple google map interface for Torch: https://github.com/ludc/googlemaps
News
- (June 2017) : Multi-view Generative Adversarial Networks pdf Mickael Chen, Ludovic Denoyer has been accepted at ECML/PKDD 2017
- (June 2017) : Two new papers:
-
- Fader Networks: Manipulating Images by Sliding Attributes
-
- Learning Time-Efficient Deep Architectures with Budgeted Super Networks
- (November 2017) Three new papers:
-
Multi-view Generative Adversarial Networks pdf Mickael Chen, Ludovic Denoyer
- NIPS Adversarial Learning workshop 2016
-
Options Discovery with Budgeted Reinforcement Learning pdf Aurelia Leon, Ludovic Denoyer
- NIPS Deep Reinforcement Learning workshop 2016
-
Modelling Relational Time Series using Gaussian Embeddings pdf Ludovic Dos Santos, Ali Ziat, Ludovic Denoyer, Benjamin Piwowarski, Patrick Gallinari
- NIPS Time Series workshop 2016
- ICONIP 2016 : Recurrent Neural Networks for Adaptive Feature Acquisition -- G Contardo, L Denoyer, T Artières -- Best student paper award
- ESANN 2016 : Policy-gradient methods for Decision Trees -- A Léon, L Denoyer
- ESANN 2016 : Learning Embeddings for Completion and Prediction of Relationnal Multivariate Time-Series -- A Ziat, G Contardo, N Baskiotis, L Denoyer
- May 2015 : New article "Reinforced Decision Trees"
- 23rd of december, 2014 : New article "Representation Learning for Cold Start Recommendation"
- November 2014: Focus on "Deep Sequential Neural Networks"
Current students
- Ali Ziat - Representation Learning models for heterogeneous temporal data (co-supervised with Nicolas Baskiotis) with VEDECOM - defense in October 2017
- Gabriella Contardo - Online Budgeted Learning (co-supervised with Thierry Artieres) - defense in july 2017
- Nassim Aklil - Budgeted approaches to Robot Navigation (co-supervised with Mehdi Khamassi) -- defense in September 2017
- Aurelia Leon - Budgeted Reinforcement Learning
- Mickael Chen - Multiview Learning (co-supervised with Thierry Artieres)
- Edouard Delasalles - Temporal Learning (co-supervised with Sylvain Lamprier)
- Guillaume Lample - Neural Machine Translation (co-supervised with MA Ranzato -- Facebook)
- Perrine Cribier-Delande - Intelligent Cars (CIFRE Renault)
Past Students
- Loreta Maag - Machine Learning approach to privacy in dynamic social networks (Alcatel)
- Gabriel Dulac-Arnold - A General Sequential Model for Constrained (Budgeted) Classification - Phd Obtained in 2014, now at DeepMind
- Yann Jacob - Machine Learning for mutilrelational and heterogeneous networks - PhD obtained in 2013 - now working at a data scientist at iRaiser
- Sheng Gao - Latent Factor Models for Link Prediction Problems - PhD obtained in 2013 - now working as a Lecturer atthe Beijing Unniversity of Post and Telecommunication
- Francis Maes - Reinforcement Learning for Structured Output Prediction - PhD obtained in 2009 - now CEO of D-Labs