Search arXiv⌕ Search

arXiv subjects

Myriam Lamolle

Publications and source records attributed to Myriam Lamolle.

3 recordsLinked to original sources

Adapting Visualization Techniques for Time-Series Anomaly Detection: From Convolutional Neural Networks to Convolutional-Recurrent Neural Networks

Deep neural networks achieve strong performance on complex tasks but are often regarded as "black boxes," which limits their adoption in domains where transparency is essential. This lack of interpretability raises ethical and legal concerns, particularly in sensitive applications such as security, where automated decisions can have serious consequences. The General Data Protection Regulation (GDPR) reinforces the need to justify decisions made by these systems. In this work, we investigate visualization techniques to improve the interpretability of anomaly detection models based on convolutional recurrent neural networks (CNN+RNN) with a TimeDistributed layer. Our architecture combines Visual Geometry Group 19 (VGG19) for feature extraction with a Gated Recurrent Unit (GRU) for sequential analysis of real-time video data. While this design is well suited for temporal inputs, the TimeDistributed layer complicates gradient propagation and weakens the link between spatial and temporal information, reducing the effectiveness of standard visualization methods. To address this challenge, we adapt techniques such as saliency maps and Gradient-weighted Class Activation Mapping (Grad-CAM) to models that incorporate a temporal dimension. Although dedicated visualization methods for such architectures remain limited, our study highlights both the difficulties and the potential of applying tools originally designed for static images to recurrent convolutional networks handling video sequences. This approach extends classical interpretation strategies to temporal models and provides an intermediate solution until more specialized methods are developed.

cs.CV↗

Towards Unveiling the Ontology Key Features Altering Reasoner Performances

Reasoning with ontologies is one of the core fields of research in Description Logics. A variety of efficient reasoner with highly optimized algorithms have been developed to allow inference tasks on expressive ontology languages such as OWL(DL). However, reasoner reported computing times have exceeded and sometimes fall behind the expected theoretical values. From an empirical perspective, it is not yet well understood, which particular aspects in the ontology are reasoner performance degrading factors. In this paper, we conducted an investigation about state of art works that attempted to portray potential correlation between reasoner empirical behaviour and particular ontological features. These works were analysed and then broken down into categories. Further, we proposed a set of ontology features covering a broad range of structural and syntactic ontology characteristics. We claim that these features are good indicators of the ontology hardness level against reasoning tasks.

cs.AI↗

Ontology Based Data Integration Over Document and Column Family Oriented NOSQL

The World Wide Web infrastructure together with its more than 2 billion users enables to store information at a rate that has never been achieved before. This is mainly due to the will of storing almost all end-user interactions performed on some web applications. In order to reply to scalability and availability constraints, many web companies involved in this process recently started to design their own data management systems. Many of them are referred to as NOSQL databases, standing for 'Not only SQL'. With their wide adoption emerges new needs and data integration is one of them. In this paper, we consider that an ontology-based representation of the information stored in a set of NOSQL sources is highly needed. The main motivation of this approach is the ability to reason on elements of the ontology and to retrieve information in an efficient and distributed manner. Our contributions are the following: (1) we analyze a set of schemaless NOSQL databases to generate local ontologies, (2) we generate a global ontology based on the discovery of correspondences between the local ontologies and finally (3) we propose a query translation solution from SPARQL to query languages of the sources. We are currently implementing our data integration solution on two popular NOSQL databases: MongoDB as a document database and Cassandra as a column family store.

cs.DB↗