Closed rfl-urbaniak closed 9 months ago
The 150 frames dataset is now in place.
I'll test the notebooks now, then plan to merge it in.
One remaining issue is that there are currently 3 copies of the foraging_toolkit folder (in collab, mx_refactors, and random_hungry_followers). I'm assuming the collab one is current, and that mx_refactored and random_hungry_followers will disappear (the relevant code will go into collab). I'll post a separate issue about this.
Yes, Collab is the current one, the rest will disappear. MX refactored should already have been deleted in this PR. No need to set separate issue, I delete as I incorporate.
In communicators_simulations.ipynb, the setup is a little confusing. Specifically, the cell starting "# custom list of locations for simulation setup 2"
Should this should be included in the cell "# Simulation setup 2 for the impact of communication"?
Also, it might be clearer to make a switch with cases (setup 1 and setup 2, with a little more explanation)
In communicators_inference, this comment (from me to myself) should be deleted, bc I already figured out that bug:
@rfl-urbaniak , should I just make this change and push?
ft.animate_birds, it'd be nice to have the legend show what the different traces are
No, it's a new comment from me to you. Check an animation based on setup 1 to see for yourself.
No, it's a new comment from me to you. Check an animation based on setup 1 to see for yourself.
Ahhh -- the previous issue was that the rewards never disappeared.
The current behavior is as intended. Each food item has a certain size, bigger than what a bird can eat in one timestep. So it shouldn't disappear immediately.
However, there's a deeper issue here, which is that the way the "calories" are defined in the code is very confusing and opens up potential bugs.
The relevant line for determining how quickly food depletes when an agent eats it is line 34 of collab/communicators/simulation.py: self.calories_acquired_per_unit_time = 5 # TO DO: make this a property of the agent, putting it here for now
This is quite hidden, and also misleading -- it appears like it could be set separately for each agent, but currently it can't. See how "delta_food_cal" is defined in the same simulation.py. I'll make an issue about this.
Yes, Collab is the current one, the rest will disappear. MX refactored should already have been deleted in this PR. No need to set separate issue, I delete as I incorporate.
Gotcha. You can resolve the issue when it's all done. I see that mx_refactored is removed in this PR -- it remains on my computer when I pull, likely due to other branches, but not on git.
collab
, with some revisions and refactoring of importsdocs
in the main folder. For now, it contains four notebooks with complete code chunks ready for review: 1)communicators_simulations.ipynb
- this is just a slight modification of the original communicators simulation notebook, renamed for uniformity 2)communicators_inference.ipynb
now contains our answer to the first communicators challenge 3) 'communicators_impact45.ipynbcontains chunks for the paper version of the second communicators problem. Might need update once we have simulations that run 150 rather than 50 frames 4)
communicators_impact30' contains code chunks with piecewise linear modeling, to be referred to but not used in the paperminor revisions to ft code, as required by the prep of the notebooks.