Emmanuel Viennet

MULTI-MODULAR NEURAL NETWORK ARCHITECTURES: APPLICATIONS IN OPTICAL CHARACTER AND HUMAN FACE RECOGNITION

In practical applications, recognition accuracy is sometimes not the only criterion; capability to reject erroneous patterns might also be needed. We show that there is a trade-off between these two properties. An efficient solution to this trade-off is brought about by the use of different algorithms implemented in various modules, i.e. multi-modular architectures. We present a general mechanism for designing and training multi-modular architectures, integrating various neural networks into a unique pattern recognition system, which is globally trained. It is possible to realize, within the system, feature extraction and recognition in successive modules which are cooperatively trained. We discuss various rejection criteria for neural networks and multi-modular architectures. We then give two examples of such systems, study their rejection capabilities and show how to use them for segmentation. In handwritten optical character recognition, our system achieves performances at state-of-the-art level, but is eight times faster. In human face recognition, our system is intended to work in the real world.

Diffusivity of the hydrogen molecule sorbed in NaA zeolite by a neutron scattering experiment

The diffusion of hydrogen in NaA zeolite was studied by incoherent neutron scattering. An experiment was carried out on samples loaded with 1.2 to 3.4 molecules per cavity and at several temperatures from 70 to 150 K. The angular (θ) dependence of the elastic and quasielastic intensities shows that the H2 molecule has a translational motion in a nonrestricted volume. A diffusion model where the molecule undergoes isotropic jumps of mean length l=3.9 A independent of temperature and is at rest for a time τ0 between two jumps accounts for the width of the quasielastic scattering in the entire (θ,T) range (τ0=10.8 ps at T=100 K). This leads to a diffusion coefficient D(cm2/s)=6×10−4 exp(E/RT) with E=2 kJ/mol for the less loaded samples. The diffusion coefficient increases slightly with the loading.

Toward the next generation of recruitment tools: an online social network-based job recommender system

This paper presents a content-based recommender system which proposes jobs to Facebook and LinkedIn users. A variant of this recommender system is currently used by Work4, a San Francisco-based software company that offers Facebook recruitment solutions. Work4 is the world leader in social recruitment technology; to use its applications, Facebook or LinkedIn users explicitly grant access to some parts of their data, and they are presented with the jobs whose descriptions are matching their profiles the most. The profile of a user contains two types of data: interactions data (users own data) and social connections data (users friends data). Furthermore the users profiles and the description of jobs are divided into several parts called fields. Our experiments suggest that to predict the users interests for jobs, using basic similarity measures together with their interactions data collected by Work4 can be improved upon. The second part of this study presents a method to estimate the importance of each field of users and jobs in the task of job recommendation. Finally, the third part is devoted to the use of a machine learning algorithm in order to improve the results obtained with similarity measures: we trained a linear SVM (Support Vector Machines). Our results show that using this supervised learning procedure increases the performance of our content-based recommender system.

Churn Prediction in a Real Online Social Network Using Local CommunIty Analysis

Prediction of user behavior in Social Networks is important for a lot of applications, ranging from marketing to social community management. In this paper, we develop and test a model to estimate the propensity of a user to stop using the social platform in a near future. This problem is called churn prediction and has been extensively studied in telecommunication networks. We focus here on building a statistical model estimating the probability that a user will leave the social network in the near future. The model is based on graph attributes extracted in the users vicinity. We present a novel algorithm to accurately detect overlapping local communities in social graphs. Our algorithm outperforms the state of the art methods and is able to deal with pathological cases which can occur in real networks. We show that using attributes computed from the local community around the user allows to build a robust statistical model to predict churn. Our ideas are tested on one of the largest French social blog platform, Sky rock, where millions of teenagers interact daily.

bitSPADE: A Lattice-based Sequential Pattern Mining Algorithm Using Bitmap Representation

Sequential pattern mining allows to discover temporal relationship between items within a database. The patterns can then be used to generate association rules. When the databases are very large, the execution speed and the memory usage of the mining algorithm become critical parameters. Previous research has focused on either one of the two parameters. In this paper, we present bitSPADE, a novel algorithm that combines the best features of SPAM, one of the fastest algorithm, and SPADE, one of the most memory efficient algorithm. Moreover, we introduce a new pruning strategy that enables bitSPADE to reach high performances. Experimental evaluations showed that bitSPADE ensures an efficient tradeoff between speed and memory usage by outperforming SPADE by both speed and memory usage factors more than 3.4 and SPAM by a memory consumption factor up to more than an order of magnitude.

A Social Formalism and Survey for Recommender Systems

This paper presents a general formalism for Recommender Systems based on Social Network Analysis. After introducing the classical categories of recommender systems, we present our Social Filtering formalism and show that it extends association rules, classical Collaborative Filtering and Social Recommendation, while providing additional possibilities. This allows us to survey the literature and illustrate the versatility of our approach on various publicly available datasets, comparing our results with the literature.

Neural nets for human face recognition

A neural network (NN) architecture based on a multilayer perceptron with shared weights is described. This kind of network allows direct gray-level image processing and lets the NN learn to extract image features in its hidden layers. These features allow fast classification of face images. The results of applying the architecture on large databases of varying difficulty, containing images taken in real-life unconstrained condition, are presented. A novel rejection criterion which allows the system to detect intruders is discussed.<<ETX>>

Field selection for job categorization and recommendation to social network users

Nowadays, in the Web 2.0 reality, one of the most challenging task for companies that aim to manage and recommend job offers is to convey this enormous amount of information in a succinct and intelligent manner such to increase the performances of matching operations against users profiles/curricula and optimize the time/space complexity of these processes. With this goal, this paper presents a novel method to formalize the textual content of job offers that aims at identifying the most relevant information and fields expressed by them and leverage this compact formalization for job recommendation and profile matching in social network environments. This method has been then developed and tested in the industrial environment represented by Multiposting and Work4, world leaders in digital solutions of e-recruitment problems. In this study three classes of documents are considered: job offers, job categories and social network user profiles (as potential job candidates); each class contains several fields with textual information. The proposed representation method permits to dynamically identify those text fields, for each class, that could help a cross-matching strategy in order to preserve, from one hand, the matching/recommendation performances and, on the other hand, reduce the cost of these operations (due to a straightforward dimensionality reduction mechanism). We then evaluated and compared the presented approach showing significant improvements on both categorization and recommendation tasks by also drastically reducing their computational costs.

Solving the Human Face Recognition Task using Neural Nets

We describe in this paper some neural network architectures designed to identify human faces from a raster image. The proposed networks are based on a multi-layer perceptron with shared weights. We discuss different hybrid architectures, combining image feature extraction by MLP and classification by specialized algorithms such as LVQ, which offer robust performances and allow the system to detect “intruders”. We present the results of our architecture on large databases of varied complexities, containing images taken in real-life unconstrained conditions.

Multiresolution scene segmentation using MLPs

The authors discuss the application of neural networks on scene segmentation problems. They demonstrate that a multiresolution image analysis followed by a scanning of smoothed images using a time delay neural network can provide solutions. This approach was applied on a task of human facial detection and localization in scenes. The system allowed the detection of 90% of the faces present in the scene. Current work on the integration of this segmentation module into a global system is discussed.<<ETX>>