An Algebraic Algorithm for the Identification of Glass Networks with Periodic Orbits along Cyclic Attractors
Open access
Date
2007-03Type
- Report
ETH Bibliography
yes
Altmetrics
Abstract
Glass piecewise linear ODE models are frequently used for simulation of neural and gene regulatory networks. Efficient computational tools for automatic synthesis of such models are highly desirable. However, the existing algorithms for the identification of desired models are limited to four-dimensional networks, and rely on numerical solutions of eigenvalue problems. We suggest a novel algebraic criterion to detect the type of the phase flow along network cyclic attractors that is based on a corollary of the Perron-Frobenius theorem. We show an application of the criterion to the analysis of bifurcations in the networks. We propose to encode the identification of models with periodic orbits along cyclic attractors as a propositional formula, and solving it using state-of-the-art SAT-based tools for real linear arithmetic. New lower bounds for the number of equivalence classes are calculated for cyclic attractors in six-dimensional networks. Experimental results indicate that the runtime of our algorithm increases slower than the size of the search space of the problem. Show more
Permanent link
https://doi.org/10.3929/ethz-a-005684186Publication status
publishedJournal / series
Technical Report / ETH Zurich, Department of Computer ScienceVolume
Publisher
Swiss Federal Institute of Technology Zurich, Department of Computer Science, Computer Systems InstituteEvent
Subject
NEURAL NETWORKS + CONNECTIONISM (ARTIFICIAL INTELLIGENCE); PROGRAMS AND ALGORITHMS FOR THE SOLUTION OF SPECIAL PROBLEMS; NEURONALE NETZWERKE + KONNEKTIONISMUS (KÜNSTLICHE INTELLIGENZ); PROGRAMME UND ALGORITHMEN ZUR LÖSUNG SPEZIELLER PROBLEMEOrganisational unit
02150 - Dep. Informatik / Dep. of Computer Science
More
Show all metadata
ETH Bibliography
yes
Altmetrics