Advanced techniques for the creation and propagation of modules in cartesian genetic programming

Abstract

The choice of an appropriate hardware representation model is key to successful evolution of digital circuits. One of the most popular models is cartesian genetic programming, which encodes an array of logic gates into a chromosome. While several smaller circuits have been successfully evolved on this model, it lacks scalability. A recent approach towards scalable hardware evolution is based on the automated creation of modules from primitive gates.In this paper, we present two novel approaches for module creation, an age-based and a cone-based technique. Further, we detail a cone-based crossover operator for use with cartesian genetic programming. We evaluate the different techniques and compare them with related work. The results show that age-based module creation is highly effective, while cone-based approaches are only beneficial for regularly structured, multiple output functions such as multipliers.

Cartesian genetic programming: its status and future

“…Noting that in modular CGP modules were acquired or destroyed randomly (i.e. via mutation), Kaufmann and Platzner introduced some new techniques for creating modules: age-based and cone-based [38]. The age-based module creation operator identifies primitives nodes that have remained unchanged for a number of generations and places these into modules (only primitive nodes can reside in a module).…”

“…Two candidates are generated using this operator and the one that is older is chosen. In contrast to standard or age-based module creation, cone-based module acquisition (MA) aggregates only primitive nodes that are within a structure called cone (see [38] for details). Cones are a widely-used concept in circuit synthesis.…”

Anomalous Diffusion within the Transcriptome as a Bio-Inspired Computing Framework for Resilience

“…As an enhancement of CGP, reusable sub-functions extend a genotype by adding more complex nodes denoted as modules [36]. The modules, which closely resemble transcripts in the biological model, are propagated through evolution by dynamically allocating and releasing them by compress and expand operators [37]. The computational operators are behaving like the transcriptome A/I shifts that expose or cover sub-functions encoded in the transcript.…”

Formal verification of candidate solutions for post-synthesis evolutionary optimization in evolvable hardware

“…In order to evolve large designs and simultaneously keep the size of chromosome small, various techniques have been proposed, including functional-level evolution [35,39], incremental evolution [43,44,45], modularization [26,51] and their combinations [12,41]. Despite the fact that a new field of computational development has attracted a lot of attention in this area and brought some theoretical as well as practical results [15,17,18,22,29,31,42,47,55] the problem of scalability is still an open issue.…”

“…The number of rows (n r (i) ) is considered as variable for a given circuit in order to represent the circuit optimally. For example, the 1408 gates of the apex1 benchmark is mapped on the array of 19x189 nodes; however only 1,5,7,14,17,26,43,57,84,117,142,177,189,187,139,89,51,27,40 gates are utilized in columns i ¼ 1. . .19.…”