- author = "Barry Porter and Penn {Faulkner Rainford} and Roberto Rodrigues-Filho",
- title = "Self-Designing Software",
- journal = "Communications of the ACM",
- year = "2025",
- volume = "68",
- number = "1",
- pages = "50--59",
- month = jan,
- keywords = "genetic algorithms, genetic programming, Genetic Improvement, Software Engineering and Programming Languages, Dana",
- ISSN = "0001-0782",
-
URL = "
https://cacm.acm.org/research/self-designing-software/",
-
DOI = "
doi:10.1145/3678165",
-
video_url = "https://vimeo.com/1040435011",
-
abstract = "Key Insights
The ability to hot-swap code at runtime, safely and with generality, allows software to continuously and autonomously reason about its own design.
When alternative implementation variants of a software component, or group of components, suit different deployment conditions, software can learn how to optimize a metric of interest by changing its design to match its environment.
By capturing short traces of function call sequences in a running system, we can replay those traces offline to automatically synthesize higher-performance implementation variants and inject those individuals into the online design pool.",
-
notes = "cites \cite{rainford:2022:GECCO}
p52 'hot swaps of code within the running system'
p55 'datacenter servers versus Raspberry Pi'
p57 'Making New Things'...GI
p58 fig 5 'Finding Better Alternatives with Genetic Improvement' Hash, cites \cite{Rainford:2022:ALife}
p58 'phylogenetic analysis in GI'
p58 'a library of diverse genetic material'
Adaptive Systems at Lancaster University, U.K.",
