Ph.D
Group : Artificial Intelligence and Inference Systems

Diagnostic distribué de systèmes respectant la confidentialité

Starts on 01/10/2007
Advisor : DAGUE, Philippe

Funding : AM
Affiliation : Université Paris-Saclay
Laboratory : LRI INRIA LEO

Defended on 27/09/2012, committee :

Research activities :

Abstract :
In this thesis, we focus on diagnosing inherently distributed systems such as peer-to-peer, where each peer has access to only a sub-part of the description of an overall system.In addition, due to a too restrictive access control policy, it can be possible that neither the peer nor the supervisor is able to explain the behaviour of the overall system.We aim at explaining this behaviour using a set of peers (each having a limited local view).
First, we show that any new system logically equivalent to the initially observed peer-to-peer setting ensures that all the diagnoses of a peer may be extended to a global diagnosis (in this case the new system ensures the correctness of the distributed diagnosis).Moreover, we prove that if the new system is structured (i.e. it contains a spanning tree for which all peers containing the same variable form a connected graph) then it ensures that any global diagnosis can be found through a set of local diagnoses (in this case the new system ensures the completeness of the distributed diagnoses).
For a succinct representation and in order to comply with the privacy policy of the vocabulary of each peer, we present a new algorithm Token Elimination (TE), which decomposes the original peer system to a structured one. We experimentally show that TE produces better quality decompositions (i.e. smaller tree widths) than other proposed methods in a distributed context.
Exploiting the structured system built by TE, we transform each local description into globally consistent DNF. We demonstrate that the latter system is correct and complete for the distributed diagnosis. Finally, we present an algorithm that can efficiently check that any local diagnosis is part of a global minimal diagnosis, turning the structured system of DNFs into a compiled system for distributed diagnosis.

CAUSAL LEARNING FOR DIAGNOSTIC SUPPORT


CAUSAL UNCERTAINTY QUANTIFICATION UNDER PARTIAL KNOWLEDGE AND LOW DATA REGIMES


MICRO VISUALIZATIONS: DESIGN AND ANALYSIS OF VISUALIZATIONS FOR SMALL DISPLAY SPACES
The topic of this habilitation is the study of very small data visualizations, micro visualizations, in display contexts that can only dedicate minimal rendering space for data representations. For several years, together with my collaborators, I have been studying human perception, interaction, and analysis with micro visualizations in multiple contexts. In this document I bring together three of my research streams related to micro visualizations: data glyphs, where my joint research focused on studying the perception of small-multiple micro visualizations, word-scale visualizations, where my joint research focused on small visualizations embedded in text-documents, and small mobile data visualizations for smartwatches or fitness trackers. I consider these types of small visualizations together under the umbrella term ``micro visualizations.'' Micro visualizations are useful in multiple visualization contexts and I have been working towards a better understanding of the complexities involved in designing and using micro visualizations. Here, I define the term micro visualization, summarize my own and other past research and design guidelines and outline several design spaces for different types of micro visualizations based on some of the work I was involved in since my PhD.