|
| 1 | +--- |
| 2 | +title: "51th regular meeting" |
| 3 | +date: 2025-11-07 |
| 4 | +--- |
| 5 | + |
| 6 | +10 ppl attended |
| 7 | + |
| 8 | +## Invited presentation |
| 9 | + |
| 10 | +- Speaker: Emily Coco |
| 11 | +- Topic: Least-cost path analysis using agent-based modelling (ABM) |
| 12 | + |
| 13 | +Emily presented her work on modelling least-cost paths on the question of Neanderthal and Denisova mobility in the pleistocene landscapes using agent-based approaches. The model restricts visibility and forward movement — agents cannot move immediatly backwards — which results in a bounded rationality movement dynamics. |
| 14 | + |
| 15 | +Originally developed in NetLogo, the model was extremely computationally demanding: a single run took around 14 days and required approximately 70 GB of storage per iteration. |
| 16 | +Emily has since ported the model to Repast Symphony, which allows more efficient execution and better control over simulation parameters. |
| 17 | + |
| 18 | +Participants raised several questions concerning the implementation and conceptual framing. Reuse and Reproducability was another focus, NASSA was proposed as model library. |
| 19 | + |
| 20 | +Emily also discussed her exploration of GPU computing, outlining current possibilities (switching to GPU and parallelisation) and open questions about how such acceleration could be effectively implemented in ABM frameworks. |
| 21 | + |
| 22 | +## Discussion highlights |
| 23 | + |
| 24 | +### Parallelisation and computational strategies |
| 25 | +- Participants exchanged ideas on the potential of parallelisation in ABM workflows. |
| 26 | +- Jim asked how Repast Symphony compares to “plain” Java implementations regarding scalability and transparency. |
| 27 | +- Not so many practical experiences seem to exist with parallelisation of ABMs, a potential aspect for further exploration |
| 28 | + |
| 29 | +### Conceptual reflections on ABM |
| 30 | +- A short discussion arose as to whether ABM is the most appropriate tool for least-cost path problems and what might be the right analytical units. |
| 31 | +- Participants considered increasing model complexity by introducing a dynamic cost surface, accounting for changing temporal and Predator-Prey dynamics. |
| 32 | + |
| 33 | +## Other CAA related activities: |
| 34 | +- Two workshops affiliated with the SSLA will take place at CAA 2026 in Vienna: |
| 35 | + - Matteo (Code Review) |
| 36 | + - Petr (Intro to R) |
| 37 | + |
| 38 | +## Decisions and actions |
| 39 | +- Maintain-a-thon initiative (Zack Batist): |
| 40 | +- Agreed to launch a “maintain-a-thon” in January 2026, where members revisit older codebases, ensure they still run, document them properly, and deposit them in long-term repositories. |
| 41 | +- The initiative will be hosted via the SSLA website, accompanied by social media outreach. |
| 42 | + |
| 43 | +## Conclusions |
| 44 | +- The maintain-a-thon was endorsed as a community activity promoting sustainability and transparency (January 2026). |
| 45 | +- GPU acceleration and parallelisation remain promising but complex paths for ABMs. |
| 46 | +- ABM continues to raise conceptual as well as technical questions about appropriate modelling scales and interpretability. |
| 47 | +- Sharing and refactoring experiences across platforms can strengthen collective learning within the community. |
| 48 | + |
| 49 | +Next SIG Meeting: December 2026, concrete date depending on the when2meet result (https://www.when2meet.com/?33002028-2srLL, details to be announced via mailing list) |
0 commit comments