Research Interests:
Research Interests:
Research Interests:
Research Interests:
One crucial tool in machine learning is a measure of partition similarity. This study focuses on the “probabilistic Rand index”, a variant of the Rand index. We look at this measure from different perspectives: probabilistic,... more
One crucial tool in machine learning is a measure of partition similarity. This study focuses on the “probabilistic Rand index”, a variant of the Rand index. We look at this measure from different perspectives: probabilistic, information-theoretic, and diversity-theoretic. These give some insight, reveal relationships with other types of measures, and suggest some possible alternative interpretations.
Research Interests:
Research Interests:
Audio signal processing is moving towards detecting and/or defining rare/anomalous sounds. The application of such an anomaly detection problem can be easily extended to audio surveillance systems. Thus, a rare sound event detection... more
Audio signal processing is moving towards detecting and/or defining rare/anomalous sounds. The application of such an anomaly detection problem can be easily extended to audio surveillance systems. Thus, a rare sound event detection method for road traffic monitoring is proposed in this paper, including detection of hazardous events, i.e., road accidents. The method is based on combining anomaly detection techniques, such as variational autoencoders (VAE) and Interval-valued fuzzy sets. The VAE is used to calculate the reconstruction error of the input audio segment. Based on this reconstruction error, a fuzzy membership function, composed of an optimistic/upper component and a pessimistic/lower component, is calculated. Finally, a probabilistic method for interval comparison is used to calculate the membership score, hence to evaluate the interval-valued fuzzy sets. Finally, classification into anomalous/normal events is obtained by defuzzification. Results show that with a careful...
Research Interests:
A large fraction of traffic on present-day Web servers is generated by bots \u2014 intelligent agents able to traverse the Web and execute various advanced tasks. Since bots\u2019 activity may raise concerns about server security and... more
A large fraction of traffic on present-day Web servers is generated by bots \u2014 intelligent agents able to traverse the Web and execute various advanced tasks. Since bots\u2019 activity may raise concerns about server security and performance, many studies have investigated traffic features discriminating bots from human visitors and developed methods for automated traffic classification. Very few previous works, however, aim at identifying bots on-the-fly, trying to classify active sessions as early as possible. This paper proposes a novel method for binary classification of streams of Web server requests in order to label each active session as \u201cbot\u201d or \u201chuman\u201d. A machine learning approach has been developed to discover traffic patterns from historical usage data. The model, built on a neural network, is used to classify each incoming HTTP request and a sequential probabilistic analysis approach is then applied to capture relationships between subsequent HTTP requests in an ongoing session to assess the likelihood of the session being generated by a bot or a human, as soon as possible. A performance evaluation study with real server traffic data confirmed the effectiveness of the proposed classifier in discriminating bots from humans at early stages of their visits, leaving very few of them undecided, with very low number of false positives
Research Interests:
Research Interests:
Research Interests:
The capacity of a clustering model can be defined as the ability to represent complex spatial data distributions. We introduce a method to quantify the capacity of an approximate spectral clustering model based on the eigenspectrum of the... more
The capacity of a clustering model can be defined as the ability to represent complex spatial data distributions. We introduce a method to quantify the capacity of an approximate spectral clustering model based on the eigenspectrum of the similarity matrix, providing the ability to measure capacity in a direct way and to estimate the most suitable model parameters. The method is tested on simple datasets and applied to a forged banknote classification problem.
Research Interests:
The feature subset task can be cast as a multiobjective discrete optimization problem. In this work, we study the search algorithm component of a feature subset selection method. We propose an algorithm based on the threshold accepting... more
The feature subset task can be cast as a multiobjective discrete optimization problem. In this work, we study the search algorithm component of a feature subset selection method. We propose an algorithm based on the threshold accepting method, extended to the multi-objective framework by an appropriate definition of the acceptance rule. The method is used in the task of identifying relevant subsets of features in a Web bot recognition problem, where automated software agents on the Web are identified by analyzing the stream of HTTP requests to a Web server.
Research Interests:
Surveillance systems are getting more and more multimodal. The availability of audio motivates a method for anomalous audio event detection (anomalous AED) for road traffic surveillance, which is proposed in this paper. The method is... more
Surveillance systems are getting more and more multimodal. The availability of audio motivates a method for anomalous audio event detection (anomalous AED) for road traffic surveillance, which is proposed in this paper. The method is based on combining anomaly detection techniques, such as reconstruction deep autoencoders and fuzzy membership functions. A baseline deep autoencoder is used to compute the reconstruction error of each audio segment. The comparison of this error to a preset threshold provides a primary estimation of outlierness. To account for the uncertainty associated to this decision-making step, a fuzzy membership function composed of an optimistic/upper component and a pessimistic/lower component is used. Evaluation results obtained after defuzzification show that with a careful parameter setting, the proposed membership function improves the performance of the baseline autoencoder for anomaly detection, and yields better or at least similar results than other anomaly detection state-of-the-art methods such as one-class SVM.
Research Interests:
In this paper we present the TARSIUS system, based on mobile technology and aimed at enhancing visually-impaired and blind people's capabilities in visual scene understanding and geolocation while are outdoor. The system components... more
In this paper we present the TARSIUS system, based on mobile technology and aimed at enhancing visually-impaired and blind people's capabilities in visual scene understanding and geolocation while are outdoor. The system components are the TARSIUS app for mobile devices, a web server, and the Remote Assistance Center. Its interface is optimized for the perceptual characteristics of its users. Moreover, the TARSIUS navigation sub-system not only leverages the GPS system, but also Bluetooth LE/iBeacon tags placed along the streets at points of interest and dangerous paths and areas.
Research Interests: Geography, Computer Science, Artificial Intelligence, Human Computer Interaction, Global Positioning System, and 8 moreGeolocation, Health and Social Science, Bluetooth, Sceince Technology Policy and Management of Innovation, Mobile Device, Industrial and Manufacturing Engineering, Visually Impaired, and Computer Networks and Communications
Research Interests:
A significant problem nowadays is detection of Web traffic generated by automatic software agents (Web bots). Some studies have dealt with this task by proposing various approaches to Web traffic classification in order to distinguish the... more
A significant problem nowadays is detection of Web traffic generated by automatic software agents (Web bots). Some studies have dealt with this task by proposing various approaches to Web traffic classification in order to distinguish the traffic stemming from human users' visits from that generated by bots. Most of previous works addressed the problem of offline bot recognition, based on available information on user sessions completed on a Web server. Very few approaches, however, have been proposed to recognize bots online, before the session completes. This paper proposes a novel approach to binary classification of a multivariate data stream incoming on a Web server, in order to recognize ongoing user sessions as generated by bots or humans. The present approach uses deep neural networks combined with Wald's Sequential Probability Ratio Test to express the relationship between subsequent HTTP requests in an ongoing session and to assess the likelihood of each session being generated by a bot or human before it ends. Experimental results showed the ability of the proposed approach to detect Web bots online with high performance scores and a small number of false negatives, as evidenced by the Recall index, minimizing the impact on human visitors. Another valuable indicator is the speed of decision: the present method allows very quick classification of nearly all sessions, leaving only very few of them undecided.
Research Interests:
Research Interests:
Audio signal processing is moving towards detecting and/or defining rare/anomalous sounds. The application of such an anomaly detection problem can be easily extended to audio surveillance systems. Thus, a rare sound event detection... more
Audio signal processing is moving towards detecting and/or defining rare/anomalous sounds. The application of such an anomaly detection problem can be easily extended to audio surveillance systems. Thus, a rare sound event detection method for road traffic monitoring is proposed in this paper, including detection of hazardous events, i.e., road accidents. The method is based on combining anomaly detection techniques, such as variational autoencoders (VAE) and Interval-valued fuzzy sets. The VAE is used to calculate the reconstruction error of the input audio segment. Based on this reconstruction error, a fuzzy membership function, composed of an optimistic/upper component and a pessimistic/lower component, is calculated. Finally, a probabilistic method for interval comparison is used to calculate the membership score, hence to evaluate the interval-valued fuzzy sets. Finally, classification into anomalous/normal events is obtained by defuzzification. Results show that with a careful parameter setting, the proposed method outperforms the state-of-the-art one-class SVM for anomaly detection.