Despite the maturity of anomaly detection in structured and image data, anomaly detection in text remains surprisingly under-explored. Existing methods struggle with two key challenges: manifold collapse in high-dimensional language embeddings, and semantic entanglement that obscures contextual anomalies. We address both with the Robust Local AutoEncoder (RLAE), which combines robust subspace recovery with local geometry preservation. It allows to learn disentangled and anomaly-sensitive representations. We extend this to RoSAE, an ensemble of randomly pruned RLAEs that enhances robustness through architectural diversity. We introduce TAC which is the first benchmark framework to distinguish anomaly types in text, using topic hierarchies to separate independent from contextual anomalies for fair and reproducible evaluation. TAC resolves inconsistencies in prior benchmarks. RoSAE consistently outperforms state-of-the-art baselines across eight diverse corpora (6 more than state-of-the-art approaches), particularly under challenging contextual settings. Our results highlight the importance of both local structure and ensemble diversity in textual anomaly detection, and position RoSAE as a robust, scalable, and efficient method for tackling both independent and contextual anomalies.

Generalized fuzzy orthopartitions extend the traditional concept of partitions to include both fuzziness and uncertainty. A generalized fuzzy orthopartition is a collection of intuitionistic fuzzy sets representing equivalence classes and satisfying a specific pair of axioms, which capture the idea that the classes must be disjoint and cover the initial universe. The aim of this article is twofold. Firstly, we aggregate and order generalized fuzzy orthopartitions by extending intuitionistic operations and relations. Secondly, we introduce and study entropy measures on generalized fuzzy orthopartitions by employing entropies on intuitionistic fuzzy sets already existing in the literature.

While fuzzy methods, and in particular fuzzy rule-based methods, have been pointed out as explainable, it is not always easy to attach a linguistic label to the conclusion provided by a rule-based system for a given observation. In this paper, we focus on the case of sparse rules, with imprecise or linguistic premises and conclusions, and their use with imprecise or linguistic observations. We explore fuzzy solutions of interpolative reasoning based on analogies, with regard to desirable mathematical properties and explainability criteria. We first recall such criteria existing in the state of the art and we analyse them in the light of explainable Artificial Intelligence (AI) requirements. We then propose a new method making easier to explain both the result of the fuzzy interpolative reasoning and the approach used to construct it. A set of experimental comparisons with some existing fuzzy interpolative reasoning approaches is presented.

In the helicopter industry, estimating the flight loads acting on mechanical components is a prerequisite for fatigue computation, and maintenance work. Currently, the flight loads assessments are done during the development phase by the manufacturer, in line with the certification guidance material specifying that maintenance intervals be set – at fleet level - to an assumed usage that is “as severe as expected in service”. This paper aims at introducing a predictive maintenance approach, based on the real usage of each helicopter, where a reliable estimation of the flight loads could help to derive the damage of the helicopter components and therefore determine when to replace them. To this end, a study is conducted to link the quasi-static evolution of the flight parameters to the dynamic evolution of the flight loads. The proposed methodology is based on the locally periodic variations of the flight loads, by applying a harmonic decomposition over each period and by extracting their real and imaginary parts. With this decomposition, different types of Machine Learning models have been compared to predict each harmonic part. To evaluate the approach, the load is finally reconstructed using the predictions and compared to the original signal. The application on real data from proprietary databases shows the relevance of the approach. As locally periodic signals are represented in many domains, this approach could therefore be useful in other fields such as healthcare or the automotive industry, where signal estimation can be crucial.

Entropies and measures of information are extensively used in several domains and applications in Artificial Intelligence. Among the original quantities from Information theory and Probability theory, a lot of extensions have been introduced to take into account fuzzy sets, intuitionistic fuzzy sets and other representation models of uncertainty and imprecision. In this paper, we propose a study of the common property of monotonicity of such measures with regard to a refinement of information, showing that the main differences between these quantities come from the diversity of orders defining such a refinement. Our aim is to propose a clarification of the concept of refinement of information and the underlying monotonicity, and to illustrate this paradigm by the utilisation of such measures in Artificial Intelligence.

Post-hoc interpretability approaches have been proven to be powerful tools to generate explanations for the predictions made by a trained black-box model. However, they create the risk of having explanations that are a result of some artifacts learned by the model instead of actual knowledge from the data. This paper focuses on the case of counterfactual explanations and asks whether the generated instances can be justified, i.e. continuously connected to some ground-truth data. We evaluate the risk of generating unjustified counterfactual examples by investigating the local neighborhoods of instances whose predictions are to be explained and show that this risk is quite high for several datasets. Furthermore, we show that most state of the art approaches do not differentiate justified from unjustified counterfactual examples, leading to less useful explanations.

Affective gaming is a hot field of research that exploits human emotion for the enhancement of player's experience during gameplay. Physiological signal is an effective modality that can provide a better understanding of the emotional states and is very promising to be applied to affective gaming. Most physiological-based affective gaming applications evaluate player's emotion on an overall game fragment. These approaches fail to capture the emotion change in the dynamic game context. In order to achieve a better understanding of psychophysiological response with a better time sensitivity, we present a study that evaluates the psychophysiological responses related to the game events. More specifically, we present a multi-modal database DAG that contains peripheral physiological signals (ECG, EDA, respiration, EMG, temperature), accelerometer signals, facial and screening recordings as well as player's self-reported eventrelated emotion assessment through game playing. We then investigate physiological-based emotion detection and recognition by using machine learning techniques. Common challenges for physiological-based affective model such as signal segmentation, feature normalization, relevant features are addressed. We also discuss factors that influence the performance of the affective models.

Improving player's game experience has always been the common goal of video game practitioner. In order to get a better understanding of player's perception of game experience, we carry out experimental study for data collection and present game experience prediction model based on machine learning method. The model is trained on the proposed multi-modal database which contains: physiological modality, behavioral modality and meta-information to predict the player game experience in terms of difficulty, immersion and amusement. By investigating the model trained on separate and fusion feature sets, we show that physiological modality is effective. Moreover, better understanding is achieved with further analysis on the most relevant features in the behavioral and meta-information features set. We argue that combining the physiological modalities with behavioral and meta information can provide a better performance on the game experience prediction.

The massive use of data mining in industrial contexts faces the problem of the results understanding by users. For example, until now, geo-scientists showdown their adoptions of these systems because of the intuitive part of their work, which requires to well understand a geological study instead of only trusting a computational system. In this paper, a new supervised machine learning based approach to build classifier from 3D-objects is presented. In this approach, first of all, an extraction of pertinent subparts of the objects is performed to highlight subparts characterising the most each kind of 3D-objects. Afterwards, a classical machine learning model is applied to build a classifier based on these pertinent subparts. The main idea is to use discriminant subparts of 3D-objects for the supervised classification in order to take care of the local nature of pertinent elements. This allows the user to be aware of these subparts which have been useful to determine the corresponding class of the object. Two algorithms are presented the 3DRESC algorithm, and its extension, the 3DRESC-TF algorithm, based on the used of transfer learning to improve its performance. Final results are presented that highlight the advantages of this new approach to build description features of 3D-objects and of the introduction of transfer learning.

This paper studies a well-established fuzzy subspace clustering paradigm and identifies a discontinuity in the produced solutions, which assigns neighbor points to different clusters and fails to identify the expected subspaces in these situations. To alleviate this drawback, a regularization term is proposed, inspired from clustering tasks for graphs such as spectral clustering. A new cost function is introduced, and a new algorithm based on an alternate optimization algorithm, called Weighted Laplacian Fuzzy Clustering, is proposed and experimentally studied.

In this paper, a comparison of the behaviour of fuzzy decision trees and gradual fuzzy decision trees is presented in a real-world application in the context of labour economics. The aim of this study is on one hand to present, in a real case, the good property of interpretability of such decision trees. On the other hand, it shows the importance to take into account a graduality relation between attributes and the class during the construction of a fuzzy decision tree. The obtained results illustrate the differences between the two types of fuzzy decision trees.

In the web of things (WOT) paradigm, it is possible for users to have access to a big amount of connected objects to fulfil their requests. However, finding the right object is a difficult task as the search should take into account not only the functionalities of the objects but also their physical localisation and their distance from the user. In this paper, a new approach to build a WOT search engine is introduced. A new semantic similarity is proposed to compare objects in ontology. To answer a user's request, the proposed model recommends objects according to both their geo-localisation and capabilities. Moreover, the search of objects takes into account the user's profile and expectations. The solution we proposed relies on fuzzy rule engines and a formal location model that characterise the search space in which relevant connected objects are selected.

In this paper a new on-line algorithm is proposed (the Droplets algorithm) for dealing with concept drifts and to produce reliable predictions. The two main characteristics of this algorithm are that it is able to adapt to different types of drifts without making any assumptions regarding their type or when they occur, and can provide reliable predictions in a non-stationary environment without using a fixed confidence threshold. Experimental results on five datasets based on Random RBF and Rotating Hyperplane generators as well as a new semi-synthetic dataset based weather temperatures show that, by discarding difficult observations, the Droplets algorithm manages to obtain the best average accuracy against ten classifiers. The results also indicate that the algorithm manages to provide reliable prediction by accurately distinguishing which observations are easily classifiable.

International audience

Photovoltaic solar panels are effective energy sources during periods of bright sunlight. Excess energy can be stored for later use at night or on cloudy days. The decision to use the stored energy now or later depends largely on being able to predict the weather on different timescales. Short term prediction of stored energy is challenging due to the non-trivial I-V characteristic of the solar cell. The erratic nature of the weather makes long term predictive energy management difficult. In this paper, we address these issues based on data collected from a solar panel, as well as its relationship to observations made of the weather. We observe that prediction, based on fuzzy decision trees, reduces the energy error by 22% compared to a constant prediction equal to the average on the studied period. Thus, exploiting the fuzzy classification provided by a fuzzy decision tree is a good improvement compared to the baseline.

In this paper, we consider the problem of the construction of fuzzy decision trees when there exists a graduality between the values of attributes and values of the class. We propose a new measure, extended from the measure of classification ambiguity, that takes into account both discrimination power and graduality with regards to the class. To highlight the importance of that kinds of measures, Medical applications is presented in which often the values of the class are symbolic and ordered and in which the discovery of gradual links between descriptive attributes and the class are seek for.

We propose an extension of the spherical K-means algorithm to deal with settings where the number of data points is largely inferior to the number of dimensions. We assume the data to lie in local and dense regions of the original space and we propose to embed each cluster into its specific ellipsoid. A new objective function is introduced, analytical solutions are derived for both the centroids and the associated ellipsoids. Furthermore, a study on the complexity of this algorithm highlights that it is of same order as the regular K-means algorithm. Results on both synthetic and real data show the efficiency of the proposed method.

In this paper, we present an original fuzzy interpolation method. In contrast to existing approaches, our method is able to always construct an interpolated fuzzy interval without a need of a special step dedicated to the "standardization" of non viable solutions, which fractures the sense of the interpolation. In fact, these "standardization" steps imply that, for instance, a point obtained from the interpolation of the upper limit (right side) of the fuzzy sets, is used to build the lower limit (left side) of the interpolated conclusion, breaking the underlying hypothesis of (linear) graduality. To achieve the direct interpolation, our method is based on the deviation of the observation from the expected linearly interpolated solution and constrains of the constructed solution between extreme cases. We illustrate and discuss the behavior of our method by comparison to other well known fuzzy interpolation methods.

In this paper, a hierarchical model of functions is presented to study and to validate functions used in an inductive learning process as measures of discrimination. This model is a fuzzy extension of a previously introduced model based on the use of any t-norm to value the intersection of fuzzy sets. Moreover, this model is based on the classically used definition of the inclusion of fuzzy sets. By means of this model, three well-known measures used to select attributes during the construction of a fuzzy decision tree are shown well-adapted as measures of discrimination. However, it is also shown that the use of an extension of the entropy of fuzzy events based on the use of Zadeh's t-norm is not convenient for such a process.

In this paper, a study is presented to explore ensembles of fuzzy decision trees. First of all, a quick recall of the state of the art related to ensembles of (fuzzy) decision trees in Machine Learning is presented. Afterwards, a new approach to construct a forest of fuzzy decision trees is proposed. Two experiments are described, one with forests of fuzzy decision trees, and the other with bagging of fuzzy decision trees. The results highlight the interest of using fuzzy set theory in this kind of approaches.

In this paper, two medical experiments are presented where the use of a fuzzy machine learning tool brought out a better understanding of the patients involved in the study. The use of fuzzy set theory to provide fuzzy labels and the construction of fuzzy decision trees to generate fuzzy rule bases enhance greatly the understandability and enable the medical scientists to have a better understanding of the correlations between the description of the patients and their medical class. The results obtained in these two experiments highlight the usefulness of fuzzy data mining approach to handle real world data and to benefit society.

On one hand, goalmouth detection is a kind of high-level semantic concept detection methods in sports video, on the other hand, data mining for video processing that can benefit video analysis is a promising research frontier. In this paper, a robust method is proposed to detect the presence of goalmouth in field-ball game video based on fuzzy decision trees. Balance process is added in training procedure. The experiment result shows that our algorithm can improve the classification when comparing with the threshold based algorithm and the decision tree based algorithm. Fuzzy rules can also be easily deduced from the constructed tree to interpret the classification model.

The construction of decision trees is an efficient tool for inductive learning, and fuzzy decision trees are particularly interesting because they enable the user to take into account imprecise descriptions of the cases, or heterogeneous values (symbolic, numerical, or fuzzy). However, since the method to construct a fuzzy decision tree is not unique, in this paper, a comparative study is presented to point out differences between three methods. This study focus on differences between methods when ranking attributes during the construction of a fuzzy decision tree. The aim is to enable the reader to understand what kind of fuzzy decision tree is obtained by each method.

In this paper, we present the discovery of rules for different challenges encountered in video indexing. These rules should be considered as knowledge that can be used as a guideline for the development of better indexing tools. We use a fuzzy decision tree to extract the rules based on color proportions of key-frames extracted from one single video-news. Experimental results and comparisons with other data mining tools are presented.

This paper is concerned with methods of construction of fuzzy decision trees and the choice of such a method the user has to make. The main differences between methods lies in the choice of the measure of discrimination that enables the ranking of attributes during the construction step. In this paper, a formal study of the main differences between measures is done. The aim is to highlight the major properties of the fuzzy decision trees constructed by a particular method.

In this paper, we propose a new algorithm to infer automatically a fuzzy partition for the universe of a set of fuzzy values, when each of these values is associated with a class. This algorithm can be used in a fuzzy decision tree system to extract knowledge from a database and to construct a set of fuzzy rules. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

We propose a new method to use an incomplete rule base with imprecise descriptions of variables. We extend classical interpolative reasoning to this case, under the assumption of graduality in variations of the variables, by using an analogical fuzzy approach.

Standard model of fuzzy computations is based on extension principle. It is known to work well, in practice, only for continuous fuzzy numbers, while producing unintuitive results when one or more arguments are discrete. It is also computationally cumbersome for all but linear operations. Another model was proposed for trapezoidal numbers only. Its operations amount to computing on the four vertices of the trapezoids, and then spanning a new trapezoid on the four resulting vertices. It is efficient, but produces fairly crude approximations for curvilinear fuzzy numbers; moreover, it is not applicable when discrete arguments are present. A model based on approximating fuzzy numbers, whether continuous or discrete, by multitrapezoidal curves and then performing coordinate-wise computations was proposed first by Ramer. It was applied to economical decision problems by his doctoral student James Wang. In this paper we place this computational method in context of fuzzy interpolations. We show how interpolation can bring quite disparate argument into a standardized form, thus permitting for efficient computations and avoid unintuitive results. Here we use the model of multiple trapezoids, but other classes of curves can be considered.

Supervised classification has already been the subject of numerous studies in the fields of Statistics, Pattern Recognition and Artificial Intelligence under various appellations which include discriminant analysis, discrimination and concept learning. Many practical applications relating to this field have been developed. New methods have appeared in recent years, due to developments concerning Neural Networks and Machine Learning. These "hybrid" approaches share one common factor in that they combine symbolic and numerical aspects. The former are characterized by the representation of knowledge, the latter by the introduction of frequencies and probabilistic criteria. In the present study, we shall present a certain number of hybrid methods, conceived (or improved) by members of the SYMENU research group. These methods issue mainly from Machine Learning and from research on Classification Trees done in Statistics, and they may also be qualified as "rule-based". They shall be compared with other more classical approaches. This comparison will be based on a detailed description of each of the twelve methods envisaged, and on the results obtained concerning the "Waveform Recognition Problem" proposed by Breiman et al.,4 which is difficult for rule based approaches.