This document gives a specification for the model used in [1]. It presents a simple way of optimizing mutual information between some input and the attractors of a (noisy) network, using a genetic algorithm. The nodes of this network are modeled as simplified versions of the structures described in the "interface theory of perception" [2]. Accordingly, the system is referred to as a "perceptual network". The present paper is an edited version of technical parts of [1] and serves as accompanying text for the Python implementation PerceptualNetworks, freely available under [3]. 1. Prentner, R., and Fields, C.. Using AI methods to Evaluate a Minimal Model for Perception. OpenPhilosophy 2019, 2, 503-524. 2. Hoffman, D. D., Prakash, C., and Singh, M.. The Interface Theory of Perception. Psychonomic Bulletin and Review 2015, 22, 1480-1506. 3. Prentner, R.. PerceptualNetworks. https://github.com/RobertPrentner/PerceptualNetworks. (accessed September 17 2020)