Looking at the very first accumulations file, I see that the there are 7286 non-zero elements in that array. It appears from reading the problem description, that 100 agents are born each turn, there are 2000 turns in total and there is no mechanism for agent to escape the environment, they have to end up somewhere inside it. Following this logic, I conclude that there should be 2e5 agents at the end of the simulation. What am I missing?
It means that if the birthplace of a new agent is occupated then it is not born.
The accumulation file should be thought of as a perfect simulation with a limited number of agents. We expect contestant submitted answers to have some agents in the accumulations, but most will be in transit within the volume. Note that we are just as interested, if not more, in how the agents move through the volume. We are looking for agent behavior that interacts with the complexity of the volume (i.e., the subsurface) and other agents.
Team @ Xeek
So is the accumulation file just the dead agents?
The accumulations could be thought of as representing dead agents and some agents that just migrated to those positions the turn before the simulation ended.
To give you some background, these accumulation files were generated from a workflow that generates synthetic seismic data (the environment file) based on several geologic inputs (faults, straigraphy, etc.). The accumulation file is a byproduct of the fluids input to this workflow, as fluids can have an influence on seismic. The fluids in this case are oil and gas.
So it’s all the dead agents and some of the live agents. That implies that the nearest neighbor calculations must have replacement. That invites a strategy of just clustering around a limited amount of accumulations.