Se connecter

Recherches récentes

Aucune recherche récente

Options de recherche

Disponible uniquement lorsque vous êtes connecté.
framapiaf.org est l'un des nombreux serveurs Mastodon indépendants que vous pouvez utiliser pour participer au fédiverse.
Un service Mastodon fourni par l'association d’éducation populaire Framasoft.

Administré par :

Statistiques du serveur :

1,4K
comptes actifs

framapiaf.org: À proposÉtatAnnuaire des profilsPolitique de confidentialité

Mastodon: À proposTélécharger l’applicationRaccourcis clavierVoir le code sourcev4.3.7

#dagstuhl

0 message0 participant0 message aujourd’hui
Public *
Aurélien Bellet

Last week, I had the privilege of discussing #privacy washing with an incredible and diverse group at #Dagstuhl. The conversations were eye-opening, and recent news keeps reminding us why this issue matters:

Ars Technica: Everything you say to your Echo will be sent to Amazon starting March 28
arstechnica.com/gadgets/2025/0

In this photo illustration, Echo Dot smart speaker with working Alexa with blue light ring seen displayed.
Ars Technica · Everything you say to your Echo will be sent to Amazon starting on March 28Par Scharon Harding
Public
Dr Vaishak Belle

#dagstuhl discussions on explainability - it is not enough to obtain an explanation (and sometimes they don’t matter to the customer, but only to the ml engineer for debugging and correctness), but also to challenge the nature of the decision. The decision rules are not set in …

Public
Dr Vaishak Belle

#Dagstuhl - how to understand semantic loss etc regularisation in terms of desired neural network distribution? Cf our recent paper using information geometry

bit.ly/3Q0aKOt

arXiv.orgSemantic Objective Functions: A distribution-aware method for adding logical constraints in deep learningIssues of safety, explainability, and efficiency are of increasing concern in learning systems deployed with hard and soft constraints. Symbolic Constrained Learning and Knowledge Distillation techniques have shown promising results in this area, by embedding and extracting knowledge, as well as providing logical constraints during neural network training. Although many frameworks exist to date, through an integration of logic and information geometry, we provide a construction and theoretical framework for these tasks that generalize many approaches. We propose a loss-based method that embeds knowledge-enforces logical constraints-into a machine learning model that outputs probability distributions. This is done by constructing a distribution from the external knowledge/logic formula and constructing a loss function as a linear combination of the original loss function with the Fisher-Rao distance or Kullback-Leibler divergence to the constraint distribution. This construction includes logical constraints in the form of propositional formulas (Boolean variables), formulas of a first-order language with finite variables over a model with compact domain (categorical and continuous variables), and in general, likely applicable to any statistical model that was pretrained with semantic information. We evaluate our method on a variety of learning tasks, including classification tasks with logic constraints, transferring knowledge from logic formulas, and knowledge distillation from general distributions.
Public
Stephan Druskat

The report of the #Dagstuhl Seminar "Research Software Engineering: Bridging Knowledge Gaps" that I co-organized with @danielskatz, Caroline Jay and Lars Grunske is out now:

> S. Druskat, L. Grunske, C. Jay, and D. S. Katz. Research Software Engineering: Bridging Knowledge Gaps (Dagstuhl Seminar 24161). In Dagstuhl Reports, Volume 14, Issue 4, pp. 42-53, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2024)
doi.org/10.4230/DagRep.14.4.42

drops.dagstuhl.deResearch Software Engineering: Bridging Knowledge Gaps (Dagstuhl Seminar 24161)
#RSEng#ResearchSoftware#SoftwareEngineering
Public
Daniel S. Katz

Our weeklong #Dagstuhl seminar called "Research Software Engineering #RSEng: Bridging Knowledge Gaps" #DagstuhlRSE (dagstuhl.de/24161) is now wrapping up

We brought together research software engineers (#RSE) and software engineering (#SE) researchers to talk about overlaps and how we can learn from each other and move forward together.

To conclude the seminar, we're talking about how we catalyze a larger community going forward, including future events, publications, videos, etc.

www.dagstuhl.deDagstuhl Seminar 24161: Research Software Engineering: Bridging Knowledge Gaps
Public
Katja Hose

I'm about to head back from another wonderful week at #Dagstuhl discussing shapes in graph data.

A big thank you to Shqiponja Ahmetaj, Slawomir Staworko, and Jan Van den Bussche for the great organization, and all participants for the engaged discussions!

And a small ironic thank you to the strikes in public transport, which kept some of us here for another day - #Dagstuhl on Friday night, a totally new experience.

dagstuhl.de/en/seminars/semina

ExplorerFlux en direct
À propos