Integrate Hotelling Law Extension into Mesa Examples #120

aumashankar · 2024-04-09T10:28:38Z

Closes issue #119: Hotelling Law Extension using Mesa (NetLego example using mesa)

Include new agents, models, and visualization modules specific to the Hotelling Law simulation.
Update requirements.txt with necessary dependencies for running the Hotelling Law model.
Ensure compatibility with existing Mesa framework and examples.
Add documentation for setting up and running the Hotelling Law model.

Online demo: http://178.128.112.142:8521/

for more information, see https://pre-commit.ci

EwoutH · 2024-04-09T12:03:49Z

Thanks a lot for the PR! Looks like an interesting model to have in our example library. I requested two of our maintainers to review it.

aumashankar · 2024-04-09T13:53:34Z

Thanks a lot for the PR! Looks like an interesting model to have in our example library. I requested two of our maintainers to review it.

Thanks a lot @EwoutH
We've developed it as a standalone, I'm seeing integration errors related to imports with build run tried to fix as I see
Looking forward to reviewer comments / feedback

rht · 2024-04-10T01:29:16Z

examples/hotelling_law/agents/agents.py

+                )
+                new_position = self.random.choice(possible_steps)
+                self.model.grid.move_agent(self, new_position)
+            elif self.model.environment_type == "line":


You could simplify the movement code by using the branch for "grid" (above), and for line environment, you simply initialize the grid with a width of 1.

@rht Please help me understand and elaborate this
You could simplify the movement code by using the branch for "grid" (above)

you simply initialize the grid with a width of 1.
It's only visualization, just wanted to show the single in center, added that way, adding width 1 it's displaying towards left with remaining grid visible, i'm not sure how it works

Need your help in explaining this

Thank you so much for the review @rht

If your concern is the centering of the visualization of the grid, it shouldn't complicate the core logic implementation. I recommend using the newer experimental visualization, https://mesa.readthedocs.io/en/stable/tutorials/visualization_tutorial.html (the current visualization you are using will be deprecated in Mesa 3.0). By default, you will get something like

This is because the size auto-adjust algorithm doesn't cover extreme case when width << height yet.
But if you pass in a specific size, e.g. 10 to the agent portrayal, you will get

I'm going through visualization with Jupyter - Does that mean visualization can only work with the Jupyter environment, and currently built-in server visualization will be deprecated? no standalone execution?
So models should be jupyter notebooks going forward...?

I'm a bit confused with API documentation and comments. I believe the current stable version of API is 2.2.4, which is what we have used in the project. Does this mean that 3.0 is an overhaul of the complete API?

Please advise if there is any document, detailing 2.2.4 vs 3.0 api, mean while I'm going through project code on mesa

gone through new UI with JupyterViz and Solara, working with new way

however, I'm facing issues with accepting params (like we have in our model, mode, Rules and mobility rate ) with choice dropdown, looks like only param it can take is no of agents with slider

@rht please help advise

All the controls in the old viz have been implemented in the new viz. See https://github.com/projectmesa/mesa/blob/4cc03a4300ce2f0600efa3ef45a61f16d2290dc0/mesa/experimental/jupyter_viz.py#L365-L371 for the checkbox. You specify as

"layout": {"type": "Select", "value": "grid", "values": ["grid", "line"]}

@rht thank you. I appreciate the patience explaining to me the required details
I've added new UI. Kindly review

examples/hotelling_law/agents/agents.py

rht · 2024-04-10T01:31:29Z

examples/hotelling_law/model/model.py

+# Function to compute the average price of all store agents in the model.
+def compute_average_price(model):
+    """Compute the average price of all stores."""
+    store_prices = [agent.price for agent in model.schedule.agents]


model.agents is the new API, and is simpler.

The whole function can be simplified to return np.mean(model.agents.get("price")).

The intermediate variables are not necessary, because the docstring is already sufficiently descriptive.

fixed these comments

examples/hotelling_law/model/model.py

rht · 2024-04-10T01:43:29Z

If you could add a test that check's for the prediction of the Hotelling's law, e.g.

mesa-examples/examples/sugarscape_g1mt/tests.py

Lines 19 to 33 in 1933cbc

    
           def test_decreasing_price_variance(): 
        
               # The variance of the average trade price should decrease over time (figure IV-3) 
        
               # See Growing Artificial Societies p. 109. 
        
               model = SugarscapeG1mt() 
        
               model.datacollector._new_model_reporter( 
        
                   "price_variance", 
        
                   lambda m: np.var( 
        
                       flatten([a.prices for a in m.schedule.agents_by_type[Trader].values()]) 
        
                   ), 
        
               ) 
        
               model.run_model(step_count=50) 
        
               df_model = model.datacollector.get_model_vars_dataframe() 
        
               check_slope(df_model.price_variance, increasing=False)

. The test is run by pytest tests.py. The functions with prefix of test_* are automatically run by pytest. Also, in the visualization, you are meant to show that the standard deviation of the price is decreasing over time.

-fixing review comments 1. The self.can_move check is redundant here, because it's already done in self.move 2. model.agents is the new API, and is simpler. 3. The whole function can be simplified to return np.mean(model.agents.get("price")). 4. The intermediate variables are not necessary, because the docstring is already sufficiently descriptive. 5. added tests for slope

for more information, see https://pre-commit.ci

aumashankar · 2024-04-11T04:17:54Z

If you could add a test that check's for the prediction of the Hotelling's law, e.g.

mesa-examples/examples/sugarscape_g1mt/tests.py

Lines 19 to 33 in 1933cbc

def test_decreasing_price_variance():

# The variance of the average trade price should decrease over time (figure IV-3)

# See Growing Artificial Societies p. 109.

model = SugarscapeG1mt()

model.datacollector._new_model_reporter(

"price_variance",

lambda m: np.var(

flatten([a.prices for a in m.schedule.agents_by_type[Trader].values()])

),

)

model.run_model(step_count=50)

df_model = model.datacollector.get_model_vars_dataframe()

check_slope(df_model.price_variance, increasing=False)

. The test is run by pytest tests.py. The functions with prefix of test_* are automatically run by pytest. Also, in the visualization, you are meant to show that the standard deviation of the price is decreasing over time.

Added tests.py

examples/hotelling_law/model.py

for more information, see https://pre-commit.ci

examples/hotelling_law/model.py

for more information, see https://pre-commit.ci

EwoutH · 2024-04-16T20:03:16Z

@aumashankar and @rht, thanks for the hard work in the past days. I'm curious, how far along the review process are we? Is there anything I can do to help?

aumashankar · 2024-04-17T01:46:18Z

@aumashankar and @rht, thanks for the hard work in the past days. I'm curious, how far along the review process are we? Is there anything I can do to help?

@EwoutH we've used old UI and some old api's with agent schedule @rht directed me to new mesa3.0 UI, helping review it in the process. last commit includes the feedback changes from @rht which can further help with review if it's good

rht · 2024-04-17T03:26:16Z

examples/hotelling_law/Readme.md

+```bash
+python run.py
+
+solara run app.py (mesa 3.0)


People might misinterpret (mesa 3.0) as something to be typed. Better omit it.

rht · 2024-04-17T03:37:22Z

examples/hotelling_law/app.py

+import solara
+from mesa.experimental import JupyterViz
+
+from .model import HotellingModel


The relative import caused this error message on my machine (Python 3.11)

ImportError: attempted relative import with no known parent package

You can remove the relative imports for now.

Adopted project structure removed relative imports

rht · 2024-04-17T03:40:25Z

I ran the Solara simulation, and observed the variance going up over time instead of going down. Doesn't this contradict Hotelling's law?

aumashankar · 2024-04-17T04:52:50Z

I ran the Solara simulation, and observed the variance going up over time instead of going down. Doesn't this contradict Hotelling's law?

This observation is due to Initial Conditions and Randomness for the basic model
we are adjusting the price with random prices taken in the range 5 -15 . The idea of this model is to show how stores will change location based on pricing and some agents will be immobile irrespective of external conditions.

Observing an increase in price variance over time in the simulation does not necessarily contradict Hotelling's Law. as it depends on factors.

We are working further with consumer preferences and market share and no of consumers (might be another PR) where

Hotelling's Law, in the context of spatial competition, suggests that businesses will choose to locate themselves in a manner that minimizes competition,
often leading to some form of clustering or minimal differentiation Hotelling's Law can imply that firms will tend to match each other's prices to avoid price wars, which in turn could lead to a reduction in price variance over time as the firms reach a sort of equilibrium state.

Edit: changed test as variance can increase or decrease depending on factors.
For now, added another adjust_price_by_neighbour_store() in agents.py to mimic some logic however, it will be fully covered with more dynamics with consumers and market share in the next PR

…ts and changed test for variance as for hotelling law variance need not decrease depends on factors

for more information, see https://pre-commit.ci

rht · 2024-04-17T11:41:10Z

This observation is due to Initial Conditions and Randomness for the basic model

The basic model itself needs to display the Hotelling law, because it is what the example is named after. At least, try changing the default params until the variance reduction happens, and in the readme, tell the reader about the param range, and when and why it starts breaking down, or that the timescale could be longer. This is standard practice for sensitivity analysis.

If you want to speed up the simulation step duration, you can tweak the param play_interval, https://github.com/projectmesa/mesa/blob/4cc03a4300ce2f0600efa3ef45a61f16d2290dc0/mesa/experimental/jupyter_viz.py#L60. Or if this doesn't do, needs to investigate why it is slow for such a low number of agents.

aumashankar · 2024-04-17T17:37:29Z

This observation is due to Initial Conditions and Randomness for the basic model

The basic model itself needs to display the Hotelling law, because it is what the example is named after. At least, try changing the default params until the variance reduction happens, and in the readme, tell the reader about the param range, and when and why it starts breaking down, or that the timescale could be longer. This is standard practice for sensitivity analysis.

If you want to speed up the simulation step duration, you can tweak the param play_interval, https://github.com/projectmesa/mesa/blob/4cc03a4300ce2f0600efa3ef45a61f16d2290dc0/mesa/experimental/jupyter_viz.py#L60. Or if this doesn't do, needs to investigate why it is slow for such a low number of agents.

There are multiple parts to it, one aspect is that, hotelling law in itself doesn't contradict as it depends on other parameters. However,
We are already working on locally with consumers and market share, I think it's better to integrate that and do sensitivity analysis.

I'll get back in a day or two to integrate those here.

…pricing and location. Removed old implementation

for more information, see https://pre-commit.ci

aumashankar · 2024-04-20T04:41:18Z

This observation is due to Initial Conditions and Randomness for the basic model

The basic model itself needs to display the Hotelling law, because it is what the example is named after. At least, try changing the default params until the variance reduction happens, and in the readme, tell the reader about the param range, and when and why it starts breaking down, or that the timescale could be longer. This is standard practice for sensitivity analysis.
If you want to speed up the simulation step duration, you can tweak the param play_interval, https://github.com/projectmesa/mesa/blob/4cc03a4300ce2f0600efa3ef45a61f16d2290dc0/mesa/experimental/jupyter_viz.py#L60. Or if this doesn't do, needs to investigate why it is slow for such a low number of agents.

There are multiple parts to it, one aspect is that, hotelling law in itself doesn't contradict as it depends on other parameters. However, We are already working on locally with consumers and market share, I think it's better to integrate that and do sensitivity analysis.

I'll get back in a day or two to integrate those here.

@rht updated PR with consumer preferences, market share, visualizations and tests for simulation

kindly review. I appreciate your patience and helping us with details (we are learning the framework)

rht · 2024-04-20T05:51:37Z

examples/hotelling_law/Readme.md

+- matplotlib
+- numpy
+- scipy
+- jupyter


jupyter is not needed. Everything else is auto-installed when installing mesa, except for scipy.

updated requirements.txt and Readme.md

Thanks a lot for the review @rht

rht · 2024-04-20T06:04:05Z

Everything else LGTM.

aumashankar · 2024-04-20T09:19:12Z

Everything else LGTM.

Thank you

rht · 2024-04-20T09:31:15Z

examples/hotelling_law/requirements.txt

This line is not needed, and is confusing to be here.

ah.. I was just going through other samples... I'm removing it... can you explain the context how other examples using it

They should be removed. The experimental visualization used to live in the mesa-examples repo.

rht · 2024-04-20T09:54:17Z

examples/hotelling_law/cprofile_test.py

@@ -0,0 +1,34 @@
+import cProfile


Forgot to comment this: the file cprofile_test.py shouldn't be part of the example.

ok. removing this file

rht · 2024-04-20T10:03:29Z

Merged, thank you @aumashankar.

aumashankar · 2024-04-20T10:21:55Z

Merged, thank you @aumashankar.

@rht Thanks a lot for the review and all the help provided in learning the mesa framework. (the old and new)
@EwoutH @jackiekazil @dmasad Thank you for the opportunity.

jackiekazil · 2024-04-20T19:09:09Z

@aumashankar Thank you for YOUR contribution and for making requested changes. We appreciate your diligence and contribution!

aumashankar and others added 5 commits April 5, 2024 20:54

Add hotelling-law project

1f6e5a0

Added team details

6b5b412

[pre-commit.ci] auto fixes from pre-commit.com hooks

4daccfc

for more information, see https://pre-commit.ci

changed import statements

486413a

fixing imports with ruff

a68b9fb

EwoutH requested review from rht and tpike3 April 9, 2024 12:01

fixing recognizing hotelling_law module by tests

e629285

jackiekazil requested review from jackiekazil and removed request for jackiekazil April 9, 2024 13:31

aumashankar added 2 commits April 9, 2024 19:11

fixing imports for recognizing hotelling_law module by tests

69099bf

fixing imports for recognizing hotelling_law module by tests

0a94531

fixing import sequence with ruff

9bd62d8

rht reviewed Apr 10, 2024

View reviewed changes

examples/hotelling_law/agents/agents.py Outdated Show resolved Hide resolved

rht reviewed Apr 10, 2024

View reviewed changes

examples/hotelling_law/model/model.py Outdated Show resolved Hide resolved

aumashankar and others added 3 commits April 11, 2024 09:16

[pre-commit.ci] auto fixes from pre-commit.com hooks

30fe6f7

for more information, see https://pre-commit.ci

- fixing feedback comment

911a63e

aumashankar added 2 commits April 11, 2024 10:18

- fixing import

e853d89

- processing feedback comment

da73a74

rht reviewed Apr 11, 2024

View reviewed changes

examples/hotelling_law/model.py Outdated Show resolved Hide resolved

aumashankar and others added 3 commits April 11, 2024 22:22

- processing feedback comment

245c763

- processing feedback comments. integrating mesa3.0 experimental UI

bad1087

[pre-commit.ci] auto fixes from pre-commit.com hooks

887c2df

for more information, see https://pre-commit.ci

rht reviewed Apr 16, 2024

View reviewed changes

examples/hotelling_law/model.py Outdated Show resolved Hide resolved

aumashankar and others added 2 commits April 16, 2024 19:54

processing feedback

46cec2c

[pre-commit.ci] auto fixes from pre-commit.com hooks

e1251fe

for more information, see https://pre-commit.ci

rht reviewed Apr 17, 2024

View reviewed changes

aumashankar and others added 3 commits April 17, 2024 10:25

Update Readme.md

3fafddc

adopted project structure with other examples, removed relative impor…

04d3eed

…ts and changed test for variance as for hotelling law variance need not decrease depends on factors

[pre-commit.ci] auto fixes from pre-commit.com hooks

b8af9e2

for more information, see https://pre-commit.ci

umashankar-xebia and others added 3 commits April 20, 2024 09:40

Added consumer preferences, market share dynamics to have changes in …

a816a8e

…pricing and location. Removed old implementation

[pre-commit.ci] auto fixes from pre-commit.com hooks

d7d0de4

for more information, see https://pre-commit.ci

fixing lint

3621d06

rht reviewed Apr 20, 2024

View reviewed changes

updated dependencies as per feedback

04fc1d3

rht reviewed Apr 20, 2024

View reviewed changes

chore: fixing dependencies

8e39a89

rht reviewed Apr 20, 2024

View reviewed changes

chore: removing profile tests

532b7bb

rht merged commit 1e0d6b8 into projectmesa:main Apr 20, 2024
3 checks passed

Integrate Hotelling Law Extension into Mesa Examples #120

Integrate Hotelling Law Extension into Mesa Examples #120

Conversation

aumashankar commented Apr 9, 2024 • edited by EwoutH Loading

EwoutH commented Apr 9, 2024

aumashankar commented Apr 9, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rht Apr 15, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rht commented Apr 10, 2024

aumashankar commented Apr 11, 2024

EwoutH commented Apr 16, 2024

aumashankar commented Apr 17, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rht commented Apr 17, 2024

aumashankar commented Apr 17, 2024 • edited Loading

rht commented Apr 17, 2024

aumashankar commented Apr 17, 2024

aumashankar commented Apr 20, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rht commented Apr 20, 2024

aumashankar commented Apr 20, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rht commented Apr 20, 2024

aumashankar commented Apr 20, 2024 • edited Loading

jackiekazil commented Apr 20, 2024

aumashankar commented Apr 9, 2024 •

edited by EwoutH

Loading

rht Apr 15, 2024 •

edited

Loading

aumashankar commented Apr 17, 2024 •

edited

Loading

aumashankar commented Apr 20, 2024 •

edited

Loading

aumashankar commented Apr 20, 2024 •

edited

Loading