- author = "Penelope {Faulkner Rainford} and Barry Porter",
- title = "Open Challenges in Genetic Improvement for Emergent Software Systems",
- booktitle = "GI @ ICSE 2021",
- year = "2021",
- month = "30 " # may,
- editor = "Justyna Petke and Bobby R. Bruce and Yu Huang and Aymeric Blot and Westley Weimer and W. B. Langdon",
- publisher = "IEEE",
- address = "internet",
- pages = "43--44",
- keywords = "genetic algorithms, genetic programming, genetic improvement",
- isbn13 = "978-1-6654-4466-8/21",
-
URL = "
https://geneticimprovementofsoftware.com/paper_pdfs/gi2021icse/rainford_gi-icse_2021.pdf",
-
video_url = "
https://www.youtube.com/watch?v=nJOXo_8P__4&list=PLI8fiFpB7BoKDaxvS7SQp0iA7fN7rrvDD&index=9",
-
video_url = "https://www.youtube.com/watch?v=rve5PbjixZw&list=PLI8fiFpB7BoKDaxvS7SQp0iA7fN7rrvDD&index=10",
-
video_url = "https://www.youtube.com/watch?v=2OE9Yiz0yMk&list=PLXTjhGKkSnI-se7uQneCX-pEDiDrQ7TIS&index=8",
-
DOI = "
doi:10.1109/GI52543.2021.00018",
- size = "2 pages",
- abstract = "Genetic improvement for emergent software systems faces unique challenges due to its deployment in highly dynamic environments. We discuss four of those challenges along with our initial plans for new research.",
-
notes = "1st author given as Penny Faulkner Rainford
'genetic improvement (GI) could automate the generation of environment-tailored variation' 'Sorting, cache replacement, or load balancing' '(a) search space coverage with limited genetic source material; (b) generalisation techniques; (c) navigating between different implementation forms; (d) leveraging human knowledge and creativity'. GitHub versus Curated material. Hash, over fitting, multiple training sets, (autotune to new language eg Japanese?) 'better select or generate GI training data'. 'evolving a shark into a dolphin', 'large leaps'. 'human guidance' 'human involvement in emergent software synthesis via GI'
How to avoid overfitting. Improved used of human guidance?
video 2OE9Yiz0yMk Penny Faulkner Rainford dont you ever forget the Battles
2:00 Discussion chair: Yu Huang
Q: benchmarks
2:42 Q: Stephanie Forrest, Q: Westley Weimer A: server management systems, adapt to dynamic traffic and load changes. Users response v energy consumption.
4:07 Q: W. B. Langdon A: transplanting genetic material between applications. Phylogenetics, horizontal gene transfer. Myra B. Cohen. A: Cyber physical systems. Self adaptive systems. Westley Weimer A: systems too dynamic, cannot precompute everything. GI taking more biological approach, eg niche finding.
7:35 Q: Yu Huang, what to tell humans so that they can guide GI? A: eg adding libraries of existing code to GI corpus
part of \cite{Petke:2021:ICSEworkshop} http://geneticimprovementofsoftware.com/events/icse2021.html",
