Learning of large-scale neural networks suffers from computational cost and the local minima problem. One solution to these difficulties is the use of modular structured networks. Proposed here is the learning of modular networks using structural learning with forgetting. It enables the formation of
The modular structure of Kauffman networks
β Scribed by U. Bastolla; G. Parisi
- Publisher
- Elsevier Science
- Year
- 1998
- Tongue
- English
- Weight
- 1003 KB
- Volume
- 115
- Category
- Article
- ISSN
- 0167-2789
No coin nor oath required. For personal study only.
β¦ Synopsis
This is the second paper of a series of two about the structural properties that influence the asymptotic dynamics of random boolean networks. Here we study the functionally independent clusters in which the relevant elements, introduced and studied in our first paper [U. Bastolla, G. Parisi, Physica D 115 (1998) 203-218], are subdivided. We show that the phase transition in random boolean networks can also be described as a percolation transition. The statistical properties of the clusters of relevant elements (that we call modules) give an insight on the scaling behavior of the attractors of the critical networks that, according to Kauffman, have a biological analogy as a model of genetic regulatory systems.
π SIMILAR VOLUMES
Kauffman networks were introduced in 1969 as a model of genetic regulatory systems. One of the most striking successes of this model is its ability to reproduce, for a critical value of its parameters, the observed scaling laws of the average cell replication time and of the average number of cell t
The state-transition matrix description of Kauffman binary networks described in the previous paper is further developed to obtain an analytical expression for the fraction of states involved in limit cycles as a function of the network size and connectivity. The result obtained for totally connecte
The sheer complexity of intracellular regulatory networks, which involve signal transducing, metabolic, and genetic circuits, hampers our ability to carry out a quantitative analysis of their functions. Here, we describe an approach that greatly simplifies this type of analysis by capitalizing on th