# zbMATH — the first resource for mathematics

Global dynamics of neural nets with infinite gain. (English) Zbl 0984.92004
Summary: We consider a model of neural and gene networks where the nonlinearities in the system of differential equations are discontinuous and piecewise constant. We develop a framework for the study of such systems. As a first step, we associate to the system a graph $$G$$ on a hypercube and show how the collection of strongly connected components of $$G$$ relates to the dynamics of the flow on the set of rays through the origin. In the second step, we discuss the relationships between the invariant sets of the ray flow and the invariant sets of the original flow. We provide a sufficient condition for a one-to-one correspondence between these sets. Finally, we study the class of binary networks within this framework. Under certain conditions, we can determine the structure of an invariant set corresponding to the lowest strongly connected component of the hypercube graph.

##### MSC:
 92B20 Neural networks for/in biological studies, artificial life and related topics 37N25 Dynamical systems in biology 68T05 Learning and adaptive systems in artificial intelligence 05C90 Applications of graph theory
Full Text:
##### References:
 [1] L. Arnold, C. Jones, K. Mischaikow, G. Raugel, Dynamical Systems Montecatini Terme 1994, Lecture Notes in Mathematics, Vol. 1609, Springer, New York, 1995. [2] Atyia, A.; Baldi, P., Oscillations and synchronizations in neural networks: an exploration of the labeling hypothesis, Int. J. neural syst., 1, 2, 103-124, (1989) [3] C.C. Conley, Isolated Invariant Sets and the Morse Index, CBMS No. 38, American Mathematical Society, Providence, RI, 1978. · Zbl 0397.34056 [4] Gedeon, T.; Fiedler, B., A class of convergent neural network dynamics, Physica D, 111, 288-294, (1998) · Zbl 0944.34039 [5] Gedeon, T.; Mischaikow, K., Structure of the global attractor of cyclic feedback systems, J. dyn. differential equations, 7, 141-190, (1995) · Zbl 0823.34057 [6] Glass, L., Combinatorial and topological methods in nonlinear chemical kinetics, J. chem. phys., 63, 4, 1325-1335, (1975) [7] Glass, L.; Pasternack, J., Stable oscillations in mathematical models of biological control systems, J. math. biol., 6, 207-223, (1978) · Zbl 0391.92001 [8] Grossberg, S., Contour enhancement, short term memory, and constancies in reverberating neural networks, Stud. appl. math., 52, 217-257, (1973) · Zbl 0281.92005 [9] Grossberg, S., Competition, decision, and consensus, J. math. anal. appl., 66, 470-493, (1978) · Zbl 0425.92017 [10] Hopfield, J.J., Neurons with graded response have collective computational properties like those of two-state neurons, Proc. natl. acad. sci., 81, 3088-3092, (1984) · Zbl 1371.92015 [11] S. Kauffman, At Home in the Universe, Oxford University Press, Oxford, 1995. [12] Mallet-Paret, J., Morse decomposition for delay-differential equations, J. differential equations, 72, 270-315, (1988) · Zbl 0648.34082 [13] Mestl, T.; Lemay, Ch.; Glass, L., Chaos in high-dimensional neural and gene networks, Physica D, 98, 33-52, (1996) · Zbl 0890.58054 [14] Mischaikow, K.; Mrozek, M.; Reineck, J., Zeta functions, periodic trajectories, and the Conley index, J. differential equations, 121, 258-292, (1995) · Zbl 0833.34045 [15] J.R. Munkres, Elements of Algebraic Topology, Addison-Wesley, Reading, MA, 1984. · Zbl 0673.55001 [16] C. Robinson, Dynamical Systems: Stability, Symbolic Dynamics, and Chaos, CRC Press, Boca Raton, FL, 1994. · Zbl 0853.58001 [17] T.W. Troyer, Feedforward Hebbian learning with nonlinear output units: a Lyapunov approach, Neural Networks 9 (2) (1996) 321-328. [18] Wazewski, T., Sur un principle topologique pour l’examen de l’allure asymptotique des integrales des equations differentiales ordinaires, Ann. soc. polon. math., 20, 279-313, (1947) · Zbl 0032.35001
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. It attempts to reflect the references listed in the original paper as accurately as possible without claiming the completeness or perfect precision of the matching.