Added homophily ratio in basic schelling example

[vbv-shm]

Changed homophily comparing logic as per discussion in #2515.
Now homophily_ratio is checked for agents to be happy. homophily_ratio is similar neighboring agents divided by the total number of neighbors.

[quaquel]

why 0.3? It would make more sense to go to 3/8 (8 is the neighborhood size of a Moore grid with a radius of 1 and all fields occupied.

[vbv-shm]

Thank you for reviewing. I will change 0.3 to 3/8 (0.375).

[vbv-shm]

@quaquel The step size of the slider is 0.1. So the ratio can either be 0.3 or 0.4.
If the ratio is 0.3, agents with 3 similar neighbors and 8 total neighbors will be happy, as 3/8 = 0.375.
And if the ratio is 0.4, agents with 3 similar neighbors and 8 total neighbors won't be happy.
So should I change 0.3 to 0.4?

[EwoutH]

Maybe we can change the step size to 0.125 as well?

[EwoutH]

Thanks for the PR! Could you update your PR description to contain all the necessary information to understand this change on its own? You can use this template if you want.

Edit: I saw there's also this PR open, which also modifies schelling:

• Fix: Schelling Model Neighbor Similarity Calculation #2518

Are there any conflicts between them? I would tend to let #2518 go first, since it was opened earlier. (or did we assign someone to this specific issue?)

[vbv-shm]

@EwoutH Thank you for the reply.
I have understood that #2815 will be given priority because it was opened first. Still, can I work on this PR just in case #2815 doesn't work out? Or is this PR required to be closed?
From now on, I will be careful not to open PR for some issue if PR is already opened.

[quaquel]

I have provided feedback in #2518, if that is not addressed in the coming days and there is no further response, I am fine with merging this instead.

[vbv-shm]

@quaquel Thank you. I understand.

[vbv-shm]

@EwoutH May you please run the benchmark tests?

[vbv-shm]

@quaquel @EwoutH Previously the benchmark tests were failing because I changed homophily to homophily_ratio in the Schelling example. homophily was provided by the configuration file while the model now takes homophily_ratio as input. In order to fix this, I changed homophily to homophily_ratio inside the configuration.
I am curious if my fix is right or not. Could you please run the benchmark tests so that changes can be checked?

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
BoltzmannWealth	small	🔵 -2.4% [-3.0%, -1.7%]	🔵 -0.8% [-1.0%, -0.6%]
BoltzmannWealth	large	🔵 -1.3% [-2.5%, -0.1%]	🔵 +0.1% [-0.7%, +0.9%]
Schelling	small	🔵 -1.0% [-1.4%, -0.7%]	🟢 -12.7% [-13.3%, -12.1%]
Schelling	large	🔵 -0.2% [-0.7%, +0.4%]	🔴 +87.2% [+85.6%, +88.6%]
WolfSheep	small	🔵 -2.6% [-3.3%, -1.9%]	🟢 -6.8% [-7.2%, -6.3%]
WolfSheep	large	🟢 -4.8% [-5.5%, -3.9%]	🟢 -8.3% [-9.8%, -6.8%]
BoidFlockers	small	🔵 -1.5% [-2.2%, -0.9%]	🔵 +2.3% [+1.5%, +3.1%]
BoidFlockers	large	🔵 -2.8% [-3.9%, -1.6%]	🔵 +1.4% [+0.7%, +2.2%]

[vbv-shm]

Looks like benchmarks ran, but performance is not up to the mark. So changing homophily to homophily_ratio inside the configuration worked.

[EwoutH]

Thanks for your work. We’re currently waiting on #2518 to be performance validated, one that’s done we can continue with rebasing this PR and checking it’s performance properly.

It’s great it now runs without problem!

[vbv-shm]

Thank you

[quaquel]

I have merged #2518, so let us know if you want to rebase this and further refine the example.

[vbv-shm]

@quaquel Yes, I would like to refine this. A few things that I think can be updated are:

1. Update configuration.py file for correct values for running benchmarks
2. Change default value of homophily from 0.3
3. Update slider step size in visualization
   As I add these changes, if something else is encountered, I will add that as well.
   Please provide your thoughts on this and anything else required to be added.

[EwoutH]

Great, make sure to git rebase first and resolve merge conflicts.

[vbv-shm]

@quaquelI @EwoutH I was running scheduling example locally, which is currently committed at the main branch (not my pull request).

No agents were getting happy, and the happy line is constant, as can be seen in the picture above.

But after I pressed the reset button, the grid shape changed, and it behaved normally, and agents are getting happy according to condition. Please see the below image:

I remember 2 weeks back when I ran it; it was running fine even without pressing reset, although the grid shape changed back then also.

[vbv-shm]

@quaquel I have added changes in my latest update, which I feel are correct. Please only see the latest update and ignore the rest before it. The changes also fix the problem that I mentioned above.

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
BoltzmannWealth	small	🔵 -0.6% [-1.4%, +0.1%]	🔵 -0.5% [-0.7%, -0.4%]
BoltzmannWealth	large	🔵 +0.4% [-0.1%, +1.0%]	🔵 -0.5% [-0.9%, -0.0%]
Schelling	small	🔵 +0.0% [-0.3%, +0.4%]	🟢 -50.9% [-51.1%, -50.8%]
Schelling	large	🔵 +1.1% [+0.6%, +1.6%]	🟢 -44.2% [-44.9%, -43.3%]
WolfSheep	small	🔵 +0.7% [+0.4%, +1.1%]	🔵 +0.3% [-0.7%, +1.4%]
WolfSheep	large	🔵 +1.1% [+0.3%, +2.0%]	🔵 +1.5% [-0.2%, +3.0%]
BoidFlockers	small	🔵 +1.2% [+0.6%, +1.8%]	🔵 +0.3% [-0.3%, +1.0%]
BoidFlockers	large	🔵 +0.8% [+0.1%, +1.4%]	🔵 +0.5% [-0.1%, +1.0%]

[quaquel]

this was 8 (the max for moore neighborhood, so it should be 1 now.

[vbv-shm]

@quaquel I will change it.

[quaquel]

one minor point remaining

[vbv-shm]

@quaquel I am assuming you are referring to the step size of the slider. Should I change slider step size from 0.1 to 0.125? Then it also makes sense to change the required homophily ratio from 0.4 to 0.375.
If these changes are made or not, 4 out of 8 similar agents are required in neighbour in order for agent to be happy. That is why I was doubtful whether to add these.

[vbv-shm]

@quaquel I have made the changes. Please check.