Skip to Main content Skip to Navigation
Theses

Modèle ARCO : Apprentissages multiples et Raisonnement réflexif sur des Connaissances hOmogènes

Abstract : In this thesis, we propose the model of an agent able to adapt to his environment. To achieve this, he discovers and efficiently uses new concepts from his experience. Knowledge is homogeneously represented in an oriented graph, so the agent can create, execute and analyse his own rules. Transmission of an activation flux in the graph allows the homogeneous and efficient use of this knowledge. The activation level defines the actual mental state of the agent, which determines the rules and the concepts that will be considered for reasoning. We present a symbolic data analysis method to induce new concepts from the graph. To create useful concepts associated with this new ones, we present although deduction rules used by the agent. Finally, control via emotions integrates new concepts by reinforcing the links and guides the agent. An agent based on this model has been developed and applied on a simple example. We were thus able to check the integrity and functionality of the model
Document type :
Theses
Complete list of metadata

https://hal.inria.fr/tel-02016467
Contributor : Philippe Caillou Connect in order to contact the contributor
Submitted on : Tuesday, August 16, 2022 - 10:26:39 PM
Last modification on : Thursday, August 18, 2022 - 3:46:08 AM

File

these4.1.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-02016467, version 1

Citation

Philippe Caillou. Modèle ARCO : Apprentissages multiples et Raisonnement réflexif sur des Connaissances hOmogènes. Intelligence artificielle [cs.AI]. Université Paris Dauphine, 2004. Français. ⟨tel-02016467⟩

Share

Metrics

Record views

70

Files downloads

5