Meetings » History » Version 1
Hannes Vilhjálmsson, 2019-04-08 12:20
| 1 | 1 | Hannes Vilhjálmsson | h1. Meeting Notes |
|---|---|---|---|
| 2 | |||
| 3 | h2. 8th of April 2019 |
||
| 4 | |||
| 5 | * READ: Work from Chris Peter's "Embodied Social Agents Lab at KTH":http://www.csc.kth.se/~chpeters/ESAL/research.html (In particular on learning and groups). |
||
| 6 | * READ: "Interesting work related to the evaluation of a crowd simulation":http://www.sbgames.org/sbgames2015/anaispdf/computacao-full/147432.pdf, based on official guidelines. Question: Where do the guidelines get their model of human behavior from? |
||
| 7 | * READ: Good source of information on how to evaluate crowd/group behavior is "work from INRIA Rennes":http://people.rennes.inria.fr/Julien.Pettre/pdf/EG2012.pdf |
||
| 8 | * QUESTION: How can we use Machine Learning (ML) to validate simulation from real world data from a specific location? |
||
| 9 | * INFO: We have a Student Innovation Fund grant for the summer, 2 students for 3 months. Project is focused on the validation of the simulation. "The Application":https://projects.cs.ru.is/attachments/download/1111/196130-0091-SimulationOfFlowAndExperience2019.pdf . |
||
| 10 | * GOAL (April/May): Set up the software and methodology for the students to execute during the summer. |
||
| 11 | * SUGGESTION (CP): We should keep an eye on the trends in AI in Europe, e.g. through reports like the "WIPO Technology Trends":https://www.wipo.int/edocs/pubdocs/en/wipo_pub_1055.pdf. |
||
| 12 | * SUGGESTION (MD): May need to formalize the environment to a greater degree to have a better chance at applying AI technologies, e.g. game environments are very well defined. |
||
| 13 | * SUGGESTION (CP): For validation the AI (ML) may really just be focused on adjusting parameters of a model that has already been constructed. I.e. tweaking the simulation/knobs to produce results closer to the real data. Not so much to alter the underlying AI of the agents. |
||
| 14 | * SUGGESTION (HV): After we have some evaluation of how close we can get with the baseline model, we continue to improve the model, with territoriality and social models getting added (CP: Realistically, bit by bit). |
||
| 15 | * SUGGESTION: We should meet with the ML people in the department, such as "Yngvi Björnsson":http://www.ru.is/faculty/yngvi/ and "Stephan Schiffel":http://www.ru.is/kennarar/stephans/cv.pdf |