Biological systems exhibit well-defined and repeatable responses, but can also explore, improvise and generate new functionality when facing unforeseen conditions. Exploratory behavior which leads to adaptation has been observed experimentally for single cells and organisms, and involves complex cellular networks such as gene regulation. However a theoretical understanding of this phenomenon is not well developed.

I will review experimental evidence for exploratory behavior in gene regulation, that stabilizes adaptive phenotypes in
single cells with high probability and relatively short times. I will then present a model of random
networks which demonstrates the capacity for such adaptation. The model addresses the problem of convergence for
random exploration in high-dimensional spaces. It shows that such convergence is non-universal and depends on network properties: it requires outgoing network hubs and is enhanced by their auto-regulation. These are  empirically validated features of gene regulatory networks; therefore their ability to support exploratory adaptation without fine-tuning, suggest a biologically plausible mechanism for primitive cellular learning.