Skip to Main content Skip to Navigation

Abstraction temporelle et interprétation quantitative/qualitative de processus à dynamiques multiples. Application aux processus biologiques

Abstract : A characteristic of complex dynamic systems, whether they are physical or natural, is the presence of processes evolving at different time-scales (multiple dynamics), and in a strongly non-linear way. We suppose that the evolution of each process under study is given by a series of measurements, or is obtained by the simulation of numerical models. To design or improve these models taking into account the scale he is interested in, the user must first be able to represent each process evolution at different time-scales, objectively determined according to the evolution itself. Then, the choice of one or several representations must be guided by an interpretation of the process dynamic features at each time-scale. We define a notion of temporal granularity, expressing the level of detail of the representation. As we deal with biological processes for which the evolution shape is more informative than precise numerical values, a quantitative/qualitative representation is proposed. A segmentation of the initial evolution in triangular episodes (expressing different shapes of local behaviors), and the successive abstraction of these triangular episodes in trapezoidal ones lead to determine every possible representation of the evolution, and every associated time-scale. The design of graphic, symbolic and statistical tools allows to automatically detect and interpret the main dynamic features of the process, at any time-scale: steadiness, stability, periodicity, recurrence, most frequent behaviours, etc. These features allow to advise the user to choose relevant time-scales, and the corresponding sampling frequencies. An additionnal functionnality identifies and locally separates the frequency components of the initial evolution. Examples of interpretation are given for two biological models. Analyses are compared with an agronomist's expert knowledge for the method validation. This work has led to the design and development of a software named PARADISE (Proces s AbstRaction AnD Interpretation SystEm).
Document type :
Complete list of metadatas
Contributor : Emilie Marchand <>
Submitted on : Tuesday, April 3, 2007 - 2:34:30 PM
Last modification on : Friday, January 10, 2020 - 9:08:08 PM
Long-term archiving on: : Saturday, May 14, 2011 - 3:38:35 AM


  • HAL Id : tel-00139748, version 1


Laurent Ayrolles. Abstraction temporelle et interprétation quantitative/qualitative de processus à dynamiques multiples. Application aux processus biologiques. Automatique / Robotique. Université Paul Sabatier - Toulouse III, 1996. Français. ⟨tel-00139748⟩



Record views


Files downloads