Diffusion scientifique
Dernières publications
Kernelization of compressing two-dimensional routing tables with order
Article dans une revue
Frédéric Giroire, Frédéric Havet, Joanna Moulierac
Given a set of communications X, which is a subset of S×T, (typically S is the set of sources and T the set of destinations), and a port function π:X→P where P is the set of ports, a routing list R is an ordered list of triples which are of the form (s,t,p), (∗,t,p), (s,∗,p) or…
Benchmarking Conventional and Learning‑Based Image Codecs for the Future JPEG DNA Standard
Communication lors d'une conférence
Claire Couvreur, Michela Testolina, Théo Ladune, Guillaume Lorand, Pierrick Philippe, Marc Antonini
In response to the ever-growing data storage demands, research has been increasingly focusing on more sustainable alternatives to traditional storage methods (e.g. HDD). One promising approach is storing cold data, i.e. rarely accessed data, on synthetic DNA, which offers…
Subtract the Corruption: Training-Data-Free Corrective Machine Unlearning using Task Arithmetic
Communication lors d'une conférence
Mostafa Mozafari, Ahmad Farooq Wani, Maria Sofia Bucarelli, Fabrizio Silvestri
Corrupted training data are ubiquitous. Corrective Machine Unlearning (CMU) seeks to remove the influence of such corruption posttraining. Prior CMU typically assumes access to identified corrupted training samples (a "forget set"). However, in many real-world scenarios the…
Uncertainty-aware knowledge representation and reasoning in critical infrastructure systems
Article dans une revue
Gérald Rocher, Nicolas Ferry, Stéphane Lavirotte, Jean-Yves Tigli
As socio-technical Cyber–Physical Systems (CPS), and Critical Infrastructure (CIs) in particular, become increasingly complex and interdependent, there is a growing need for integrated approaches that support decision-making across a wide range of domains. These domains rely on…
Assessing Error-Robustness of JPEG Codecs for DNA Image Storage
Communication lors d'une conférence
Michela Testolina, Claire Couvreur, Marc Antonini
The exponential growth of digital data motivates the exploration of sustainable alternatives to conventional storage. Synthetic DNA is a promising medium for long-term data retention due to its ultra-high storage density, low energy requirements, and long-term durability…
Beyond Image-Level Evaluation: Specimen-Aware Multi-View Feature Fusion for Wood Charcoal Classification
Communication lors d'une conférence
Dieu-Donné Fangnon, Diane Lingrand, Marco Corneli, Isabelle Théry-Parisot, Antoine Pasqualini
Automated identification of wood species from microscopic images of charcoal is a key challenge in anthracology, with applications in archaeology, ecology, and forest heritage studies. Existing deep learning approaches for this task suffer from inflated accuracy estimates…