Francoise Soulie Fogelman

Hub France IA

Learning, defined as the process of constructing meaning and developing competencies to act on it, is instrumental in helping individuals, communities, and organizations tackle challenges. When these challenges increase in complexity and require domain knowledge from diverse areas of expertise, it becomes difficult for single individuals to address...

This chapter summarizes contributions made by Ricardo Baeza-Yates, Francesco Bonchi, Kate Crawford, Laurence Devillers and Eric Salobir in the session chaired by Françoise Fogelman-Soulié on AI & Human values at the Global Forum on AI for Humanity. It provides an overview of key concepts and definitions relevant for the study of inequalities and Ar...

The use of artificial intelligence (AI) in a variety of research fields is speeding up multiple digital revolutions, from shifting paradigms in healthcare, precision medicine and wearable sensing, to public services and education offered to the masses around the world, to future cities made optimally efficient by autonomous driving. When a revoluti...

Community detection in networks is a fundamental data analysis task. Recently, researchers have tried to improve its performance by exploiting semantic contents and interpret the communities. However, they typically assume that communities are assortative (i.e. vertices are mostly connected to others within the group), thus they cannot find the gen...

In-memory data processing frameworks (e.g., Spark) make big data analysis greatly simpler and efficient. However, stragglers that take much longer to finish than other tasks significantly degrade performance. There exist multiple factors that cause stragglers, either from the hardware resource layer or application layer, e.g. hardware heterogeneity...

Feature Engineering (FE) is one of the most beneficial, yet most difficult and time-consuming tasks of machine learning projects, and requires strong expert knowledge. It is thus significant to design generalized ways to perform FE. The primary difficulties arise from the multiform information to consider, the potentially infinite number of possibl...

This chapter discusses the problem of attributed graph clustering when the vertices are described by real attributes. Community detection deals with the unsupervised clustering of graph vertices in social networks. Experiments on synthetic datasets aim at evaluating the performances of I‐Louvain, which exploits attributes and relational data on art...

Online experience is systematically enhanced through recommender systems (RSs) bringing to users recommendations for movies, products, songs, friends, banners or content on social sites or travels. This chapter introduces RSs and the major algorithms to produce recommendations. It presents social networks and their properties and discusses the part...

Finding semantic communities using network topology and contents together is a hot topic in community detection. Existing methods often use word attributes in an indiscriminate way to help finding communities. Through the analysis we find that, words in networked contents often embody a hierarchical semantic structure. Some words reflect a backgrou...

Recent research on community detection focuses on learning representations of nodes using different network embedding methods, and then feeding them as normal features to clustering algorithms. However, we find that though one may have good results by direct clustering based on such network embedding features, there is ample room for improvement. M...

Topic models have many important applications in fields such as Natural Language Processing. Topic embedding modelling aims at introducing word and topic embeddings into topic models to describe correlations between topics. Existing topic embedding methods use documents alone, which suffer from the topical fuzziness problem brought by the introduct...

We present a framework called Learning Automatic Feature Engineering Machine (LAFEM), which formalizes the Feature Engineering (FE) problem as an optimization problem over a Heterogeneous Transformation Graph (HTG). We propose a Deep Q-learning on HTG to support efficient learning of fine-grained and generalized FE policies that can transfer knowle...

Recent research on community detection focuses on learning representations of nodes using different network embedding methods, and then feeding them as normal features to clustering algorithms. However, we find that though one may have good results by direct clustering based on such network embedding features, there is ample room for improvement. M...

Feature engineering is one of the most difficult and time-consuming tasks in data mining projects, and requires strong expert knowledge. Existing feature engineering techniques tend to use limited numbers of simple feature transformation methods and validate on simple datasets (small volume, simple structure), obviously limiting the benefits of fea...

Community detection is one of the most important tasks in network analysis. Many community detection methods have been proposed recently. However, they typically focus on assortative community structures (i.e. nodes within the same community have more connections), while ignoring the diversity of community patterns in real world. In addition, the n...

Using network topology and semantic contents to find topic-related communities is a new trend in the field of community detection. By analyzing texts in social networks, we find that topics in networked contents are often hierarchical. In most cases, they have a two-level semantic structure with general and specialized topics, to respectively denot...

Service clustering is the foundation of service discovery, recommendation and composition. Most of the existing methods mainly use service attribute information and ignore the semantic-based invocation relationships among service users. In fact, mutual invocation relationships between services occur on operations of the corresponding services, whil...

Recommendation is widely used in our daily life. Especially in the e-commerce area, a good recommendation system can help users a lot. In this paper, we introduce our approach for the KKbox's Music Recommendation Challenge. In this challenge, we were asked to build a recommendation system that can predict whether a user will listen again to a song...

Due to the importance of community structure in understanding network and a surge of interest aroused on community detectability, how to improve the community identification performance with pairwise prior information becomes a hot topic. However, most existing semi-supervised community detection algorithms only focus on improving the accuracy but...

Discovery of communities in networks is a fundamental data analysis task. Recently, researchers have tried to improve its performance by exploiting node contents, and further interpret the communities using the derived semantics. However, the existing methods typically assume that the communities are assortative (i.e. members of each group are most...

Community detection is an important task in social network analysis. Existing methods typically use the topological information alone, and ignore the rich information available in the content data. Recently, some researchers have noticed that user profiles can also benefit to community detection, and hence the combination of topology and node conte...

Variety is an important factor for obtaining benefits expected from Big Data. Even with rich initial data sets, one should always increase variety by generating new attributes. One will obtain simpler and more performant models.

In this paper (in french), we introduce major concepts for Social Network analysis. Without getting into technical details (covered in the many references provided), we illustrate these concepts through results obtained in various collaborative research projects in which we participated.

Big Data analytics present both opportunities and challenges for companies. It is important that, before embarking on a Big Data project, companies understand the value offered by Big Data and the processes needed to extract it. This chapter discusses why companies should progressively increase their data volumes and the process to follow for imple...

Social media and more generally online social networking technologies have emerged as powerful tools to exchange information among a large variety of players, including the public, authorities, companies and journalists. In this paper, we review the present and potential uses of social media and how to value information they contain to manage secur...

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 possibiliti...

We present recommendation techniques used in AMMICO, an on-going project to develop smart audio-guides for museums. We propose several recommendation approaches relying on social network analysis and discuss how such techniques could address museum-specific issues and enhance museum visitors’ experience. We present a prototype of such a technique w...

The method uses predictive analysis to determine a model based on past data including a first social network built between communicating entities for a first observation period and behavioral centrality measures derived from behavioral data observed in a following time period. The model thus determined is then applied to a second social network bui...

Avec la participation de François Bancilhon (Data publica) François Bourdoncle (Dassault systèmes) Stephan Clemencon (Telecom ParisTech) Colin de la Higuera (U. Nantes, SIF) Gilbert Saporta (CNAM) Francoise Soulie-­‐Fogelman (Kxen) François Bourdoncle et Paul Hermelin ont été nommés « chefs de file » de la filière Big Data française. Leur mission e...

Large social networks provide many services to their users for free. To be able to do so, these sites need to monetize their audience, in order to increase the level of services offered to their users and develop their business. Monetization of a social network platform comes not only from efficient advertising features but also from all features w...

The method uses predictive analysis to determine a model based on past data including a first social network built between communicating entities for a first observation period and behavioral centrality measures derived from behavioral data observed in a following time period. The model thus determined is then applied to a second social network bui...

Collaborative filtering has been extensively studied in the context of ratings prediction. However, industrial recommender systems often aim at predicting a few items of immediate interest to the user, typically products that (s)he is likely to buy in the near future. In a collaborative filtering setting, the prediction may be based on the user's p...

The online services industry is a rapidly growing industry with a worldwide online ad market projected to grow from $48 billion in 2011 to $67 billion in 2013, of which 47% will come from display advertising and 53% from search advertising. Online Services ...

Résumé. Le data mining est aujourd'hui de plus en plus utilisé dans les entreprises les plus compétitives. Ce développement, rendu possible par la disponibilité grandissante de masses de données importantes, pose des contraintes tant théoriques (quels algorithmes utiliser pour produire des modèles d'analyses exploitant des milliers de variables pou...

Business Intelligence is a very active sector in all industrial domains. Classical techniques (reporting and Olap), mainly concerned with presenting data, are already widely deployed. Meanwhile, Data Mining has long been used in companies as a nichetechnique, reserved for experts only and for very specific problems (credit scoring, fraud detection...

We present the problem of « Reject Inference » for credit acceptance. Because of the current legal framework (Basel II), credit institutions need to industrialize their processes for credit acceptance, including Reject Inference. We present here a methodology to compare various techniques of Reject Inference and show that it is necessary, in the ab...

Vapnik’s work on learning theory has provided a general framework for building efficient predictors. Through the notion of Vapnik-Chervonenkis dimension, we can control the predictor complexity and adapt it to the problem complexity and sample size, while guaranteeing a worst case error bound. In this talk, I will focus on practical issues to see h...

this paper, we will talk only about OCR applications: while our approach is general to image processing, OCR serves as a perfect test bed for new algorithms, because of the extensive litterature and "universal" data bases (the NIST data base) available. and also of how big the learning set will have to be (i.e. how many data will be needed to match...

There are several ways to design neural networks so that they would generalize better. Specifying an architecture with local connections is one of these, but the choice for the size and location of the neighborhoods involved in this method always seems to be ad hoc. We propose to add an extra term to the error function, which forces a fully connect...

We show in this paper how Neural Networks can be used for Human Face Processing. In Part I, we show how Neural Networks can be viewed as a particular class of Statistical models. We introduce learning as an estimation problem 1, then describe Multi-Layer Perceptrons and Radial Basis Function networks 2, widely used Neural Networks which we will use...

The authors introduce the conventional multilayer perceptron and Vapnik's (1995) theory of local estimation: results comparing these techniques are given. They then present two applications of prediction: one in a check processing center, and the other in a credit card call center. Both applications are presently operating at sites run by SLIGOS. F...

In this paper, we present 3 different neural network-based methods to perform variable selection. OCD — Optimal Cell Damage — is a pruning method, which evaluates the usefulness of a variable and prunes the least useful ones (it is related to the Optimal Brain Damage method of Le Cun et al.). Regularization theory proposes to constrain estimators b...

this paper we will show how NNs can be used for variable selection with a criterion based upon the evaluation of a variable usefulness. Various methods have been proposed to assess the value of a weight (e.g. saliency [Le Cun et al. 90] in the Optimal Brain-Damage -OBD- procedure): along similar ideas, we derive a method, called Optimal Cell Damage...

Sigmoid-like activation functions implemented in analog hardware differ in various ways from the standard sigmoidal function as they are asymmetric, truncated, and have a non-standard gain. It is demonstrated how one can adapt the backpropagation learning rule to compensate for these non-standard sigmoids as available in hardware. This method is ap...

It is argued that maximal benefit is gained from complex Neural Network architectures. A formalism to train Multi Modular Architectures is given. Examples of such MMAs for various applications in image and speech processing, and time series prediction are shown, which illustrate the benefits of using MMAs.

A formalism to train multi-modular architectures (MMAs) is given. Examples of MMAs for various applications in image and speech processing are shown, which illustrate the benefits of using MMAs.

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-modu...

Special issue on Advances in Pattern recognition using neural network technologies. I. Guyon, P. Wang, eds.

The work presented here describes a method for the development of systems for the recognition of 3-D objects for which 2-D plans or technical drawings are available. CAD tools are used to generate 3-D views of an object from its 2-D plan, thus providing a learning database for a neural network. Tests are performed on the recognition of electricity...

We discuss here the application of Neural Networks to a human face segmentation task. We demonstrate that a multiresolution image analysis followed by a scanning of smoothed images using a Time Delay Neural Network can provide solutions. We discuss the results obtained using a recognition network and a two class network. The performances of the sys...

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...

In this paper we describe one way to solve the real world optical character recognition problem. We test several multilayer perceptron architectures, using strong constraints and shared weights and show that the cooperation between the GBP and LVQ algorithms allows to reach better performances on real world databases than “classical” techniques. We...

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...

One way to solve the real-world optical character recognition (OCR) problems is described. The strategy chosen was to introduce a neural character recognition box into a classical OCR product. The authors recall the different steps involved in the OCR process. Some of the problems arising in the design of a database for the training of neural netwo...

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 database...

Neural networks can be used in many different areas of problems related to petroleum exploration and production. There already exist well defined classes of applications, together with appropriate neural networks architectures. Detailed theoretical results allow us to monitor and evaluate the results obtained by neural networks. Sophisticated appli...

Neural Networks are very efficient for real world applications. However, practical problems often arise which can hinder performances. We discuss here some of these problems: under-representation of classes, rejection of outliers and ambiguous patterns, and illustrate the issues raised through various applications.