Julien Perez

Julien Perez
Naver Labs Europe · Machine Learning and Optimization group

PhD

About

44
Publications
5,081
Reads
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
574
Citations
Additional affiliations
September 2011 - present
Université Paris-Est Créteil Val de Marne - Université Paris 12
Position
  • PostDoc Position
September 2011 - September 2013
Université Paris-Est Créteil Val de Marne - Université Paris 12
Position
  • Lecturer - Analog electronic - Signal processing
September 2010 - September 2011
MINES ParisTech
Position
  • PostDoc Position

Publications

Publications (44)
Conference Paper
Full-text available
Deploying embodied agents in real-world settings brings safety, adaptability, and costs challenges. As Reinforcement Learning (RL) in robotics remains expensive in such scenarios, generic decision models are often trained at a high scale of simulation. Then, transferring to a real-world environment becomes the starting point of a lifelong adaptatio...
Conference Paper
Training agents to operate in one environment often yields overfitted models that are unable to generalize to the changes in that environment. However, due to the numerous variations that can occur in the real-world, the agent is often required to be robust in order to be useful. This has not been the case for agents trained with reinforcement lear...
Article
Full-text available
Partially observable environments present an important open challenge in the domain of sequential control learning with delayed rewards. Despite numerous attempts during the two last decades, the majority of reinforcement learning algorithms and associated approximate models, applied to this context, still assume Markovian state transitions. In thi...
Conference Paper
We report in this paper our contribution to the FEIII 2017 challenge addressing relevance ranking of passages extracted from 10-K and 10-Q regulatory filings. We leveraged our previous work on document structure and content analysis for regulatory filings to train hybrid text analytics and decision making models. We designed and trained several lay...
Conference Paper
Full-text available
Article
Full-text available
Many methods have been used to recognize author personality traits from text, typically combining linguistic feature engineering with shallow learning models, e.g. linear regression or Support Vector Machines. This work uses deep-learning-based models and atomic features of text, the characters, to build hierarchical, vectorial word and sentence re...
Article
Full-text available
Machine reading using differentiable reasoning models has recently shown remarkable progress. In this context, End-to-End trainable Memory Networks, MemN2N, have demonstrated promising performance on simple natural language based reasoning tasks such as factual reasoning and basic deduction. However, other tasks, namely multi-fact question-answerin...
Article
In an end-to-end dialog system, the aim of dialog state tracking is to accurately estimate a compact representation of the current dialog status from a sequence of noisy observations produced by the speech recognition and the natural language understanding modules. A state tracking module is primarily meant to act as support for a dialog policy but...
Article
In an end-to-end dialog system, the aim of dialog state tracking is to accurately estimate a compact representation of the current dialog status from a sequence of noisy observations produced by the speech recognition and the natural language understanding modules. A state tracking module is primarily meant to act as support for a dialog policy but...
Article
In an end-to-end dialog system, the aim of dialog state tracking is to accurately estimate a compact representation of the current dialog status from a sequence of noisy observations produced by the speech recognition and the natural language understanding modules. A state tracking module is primarily meant to act as support for a dialog policy but...
Article
In an end-to-end dialog system, the aim of dialog state tracking is to accurately estimate a compact representation of the current dialog status from a sequence of noisy observations produced by the speech recognition and the natural language understanding modules. A state tracking module is primarily meant to act as support for a dialog policy but...
Conference Paper
Full-text available
This paper presents our contribution to the Se-mEval 2016 task 5: Aspect-Based Sentiment Analysis. We have addressed Subtask 1 for the restaurant domain, in English and French, which implies opinion target expression detection , aspect category and polarity classification. We describe the different components of the system, based on composite model...
Article
Full-text available
The task of dialog management is commonly decomposed into two sequential subtasks: dialog state tracking and dialog policy learning. In an end-to-end dialog system, the aim of dialog state tracking is to accurately estimate the true dialog state from noisy observations produced by the speech recognition and the natural language understanding module...
Article
This paper presents two approaches for energy management in the context of mobile sensor networks. Such networks are characterised by well-studied energy consumption models where the discovery procedure (a.k.a. 'probing phase') of a majority of transmission protocols is extremely expensive, as compared to other network operations. Optimising the pr...
Conference Paper
Full-text available
This technical notebook describes the methodology used – and results achieved – for the PAN 2015 Author Profiling Challenge by the team from Xerox Research Centre Europe (XRCE). This year, personality traits are introduced alongside age and gender in a corpus of tweets in four languages – English, Spanish, Italian and Dutch. We describe a largely l...
Article
Full-text available
Stochastic Gradient Descent (SGD) is one of the most widely used techniques for online optimization in machine learning. In this work, we accelerate SGD by adaptively learning how to sample the most useful training examples at each time step. First, we show that SGD can be used to learn the best possible sampling distribution of an importance sampl...
Conference Paper
Full-text available
Language use is known to be influenced by personality traits as well as by socio-demographic characteristics such as age or mother tongue. As a result, it is possible to automatically identify these traits of the author from her texts. It has recently been shown that knowledge of such dimensions can improve performance in NLP tasks such as topic an...
Conference Paper
In the last decade we’ve seen advances in speech recognition, natural language understanding, natural language generation, and speech synthesis to such an extent that conversational interfaces are becoming possible. Indeed, personal assistants like Apple Siri, Microsoft Cortana, and Nuance DMA have brought conversational agents into popular use. In...
Conference Paper
A main objective of Information Centric Network (ICN) is to improve the network by placing the knowledge in center of network design. This vision of the network needs an efficient distributed and decentralized knowledge plane. So, an important amount of knowledge should be disseminated over the supervised network, which remains an open problem. Ind...
Conference Paper
Full-text available
A main objective of an Information Centric Network (ICN) is to improve the network by placing the knowledge in center of the network design. This vision of the network needs an efficient distributed and decentralized knowledge plane. So, an important amount of knowledge should be disseminated over the supervised network, which remains an open probl...
Conference Paper
Full-text available
In this paper, a novel approach of energy management in the context of mobile sensor networks is presented. In such a context, no guarantee can be assumed that a fully connected path between any sources and destinations exists at any time. Consequently, the discovery procedure, also known as probing phase, of any mobile wireless transmission infras...
Conference Paper
Full-text available
The question of objective measurement of quality of images remains an opened issue. In this paper, we conduced a study concerning the direct measurement of perception of image quality using frontal electroencephalography (EEG). In our work, subjects viewed a series of images for a short predefined period of time while their brain activity was regis...
Article
Full-text available
Grids organize resource sharing, a fundamental requirement of large scientific collaborations. Seamless integration of grids into everyday use requires responsiveness, which can be provided by elastic Clouds, in the Infrastructure as a Service (IaaS) paradigm. This paper proposes a model-free resource provisioning strategy supporting both requireme...
Article
As current server capacity and network bandwidth become increasingly overloaded by the rapid growth of high quality emerging multimedia services such as mobile online gaming, social networking or IPTV, a critical factor of success of these multimedia services becomes the end-user perception of quality while them using the service. As a result, user...
Conference Paper
Nowadays, the end-users have multitude of offers that make service prices decrease in using a network environment with multiple-operators and multiple-network. Consequently, the competition between network providers increases. The customer is today in a strong position with ability to select the best one among different competing providers. Besides...
Article
Full-text available
Résumé : Two recurrent questions often appear when solving numerous real world policy search problems. First, the variables defining the so called Markov Decision Process are often continuous, that leads to the necessity for discretization of the considered state/action space or the use of a regression model, often non-linear, to approach the Q-fun...
Conference Paper
Full-text available
We used in the past a lot of computational power and human expertise for having a very big dataset of good 9x9 Go games, in order to build an opening book. We improved a lot the algorithm used for gen- erating these games. Unfortunately, the results were not very robust, as (i) opening books are definitely not transitive, making the non-regression...
Conference Paper
Full-text available
Modern game playing programs use opening books in order to perform better. Generating opening books automatically in combination with an alpha-beta program has been well studied. A challenge is to generate automatically an opening book for the new Monte-Carlo Tree Search (MCTS) algorithms. In this article, we tackle this issue by combin- ing two le...
Conference Paper
Full-text available
The Affinity Propagation (AP) clustering algorithm proposed by Frey and Dueck (2007) provides an understandable, nearly optimal summary of a dataset, albeit with quadratic computational complexity. This paper, motivated by Autonomic Computing, extends AP to the data streaming framework. Firstly a hierarchical strategy is used to reduce the complexi...
Article
Full-text available
Two production models are candidates for e-science computing: grids enable hardware and software sharing; clouds propose dynamic resource provisioning (elastic computing). Organized sharing is a fundamental requirement for large scientic collaborations; responsiveness, the ability to provide good response time, is a fundamental requirement for seam...
Conference Paper
Full-text available
This paper presents a successful application of parallel (grid) coevolution applied to the building of an opening book (OB) in 9x9 Go. Known sayings around the game of Go are refound by the algorithm, and the resulting program was also able to credibly comment openings in professional games of 9x9 Go. Interestingly, beyond the application to the ga...
Article
Full-text available
Nous combinons pour de l'exploration Monte-Carlo d'arbres de l'apprentissage arti- RÉSUMÉ. ficiel à 4 échelles de temps : – regret en ligne, via l'utilisation d'algorithmes de bandit et d'estimateurs Monte-Carlo ; – de l'apprentissage transient, via l'utilisation d'estimateur rapide de Q-fonction (RAVE, pour Rapid Action Value Estimate) qui sont app...
Conference Paper
Full-text available
The main contribution of this paper is the presentation of a general scheduling framework for providing both QoS and fair-share in an autonomic fashion, based on 1) configurable utility functions and 2) RL as a model-free policy enactor. The main difference in our work is that we consider a multi-criteria optimization problem, including a fair-shar...
Article
Full-text available
Nous ajoutons différentes astuces d'expertise Go dans un programmation de planification Monte-Carlo à partir de bandits, via des simulations virtuelles ajoutées aux statistiques de bandits.
Conference Paper
Full-text available
Large scale production grids are a major case for autonomic computing. Following the classical definition of Kephart, an autonomic computing system should optimize its own behavior in accordance with high level guidance from humans. This central tenet of this paper is that the combination of utility functions and reinforcement learning (RL) can pro...
Article
Full-text available
Large scale production grids are an important case for autonomic computing. They follow a mutualization paradigm: decision-making (human or automatic) is distributed and largely independent, and, at the same time, it must implement the highlevel goals of the grid management. This paper deals with the scheduling problem with two partially conflictin...
Patent
Full-text available
We combine for Monte-Carlo exploration machine learning at four dierent time scales: - online regret, through the use of bandit algorithms and Monte-Carlo estimates; - transient learning, through the use of rapid action value estimates (RAVE) which are learnt online and used for accelerating the explo- ration and are thereafter neglected; - oine le...
Article
Full-text available
Predicting the performance of schedulers is a notoriously difficult task. As a consequence, grid users might be tempted to work around the standard grid middleware by designing specific strategies, which would be counterproductive if generally adopted. On the other hand, Machine Learning has been successfully applied to performance prediction in di...

Network

Cited By