Complexity Explorer Tutorial

[tpike3]

Team: Its Official!

We are building a Mesa Tutorial to be hosted by Santa Fe Institute's Complexity Explorer.

We will use this discussion to bring it all together based on our many other disparate conversations.

Linked is our proposal and the Complexity Explorer guidelines.

If I was going to summarize the key points of our talk and guidelines, Complexity Explorer is easy to work with and very supportive. We should look to make 12 minute or less videos and keep them direct and on point due to optimize effectiveness for its global audience.

Initial Timeline: By June 15th send Complexity Explorer first 2 videos. Intro (discussing ABMs) and Part 1 of Mesa tutorial

Question 1: What ABM are we going to build?
Question 2: What environment should we use?

[rht]

Question 1: I'm voting for Sugarscape because it is classic, and has multiple variations.
Question 2: Google Colab seems to be more responsive than Binder.

[jackiekazil]

Agree

[tpike3]

Sugarscape and colab it is

[tpike3]

I have been chewing on this and it seems the intro part set up mesa on colab etc should be pretty easy.

However for the 12 minute or less intro what do we want to focus on ---

I think the standard why ABMs, a quick overview of Mesa as an open source, everyone welcome project and then overview of the tutorial plan?

[jackiekazil]

What do want people to walk away with in 12 minutes? (Think 12 slides or less.)

[tpike3]

I would say:

• ABMs are an emerging scientific tool to explore complex adaptive system based phenomenon.
• Mesa is a global open source project to advance this tool and all our welcome
• This tutorial will:
  • Introductory: walk users through building the Sugarscape (traders) models
  • Advanced: how to make contributions
  • Future ones will be produced as needed (i.e. Mesa-Geo)

[rht]

We don't have Sugarscape trader model (G1MT in the example yet. We only have the constant growback version.

[rht]

I do have the G1MT code on my local machine, but it's not fully tested yet.

[tpike3]

I am starting to code out the model and for a June deadline of our initial videos to SFI to get feedback I am thinking:

1 - ABMs as a tool (do at the end)
2 - Overview of the sugarscape traders model (do first and submit as just overview talking material)
3 - part 1 of ??? (do second and submit as coding material)

The million dollar question particularly for 2 is who wants to be the face of Mesa (@jackiekazil) ??).

[jackiekazil]

is this part 1, part 2, part 3? or 3 parts to one 12 minute chunk?

[tpike3]

3 separate videos which are 12 minutes or less so pat 1, part 2, part 3

[tpike3]

I have started the colab "script" for the tutorial. The idea is to have the code for each section that the presenter can just type in and notes. Like writing the idea is to do it iteratively and constantly refine.

Code -----

In the resources file is the JSON, plus some visual adds. Currently they are a picture of sugarscape from mesa interface, a pic with all the numbers and a pic zoomed in on a hill --- the idea is for the visual aids to be discussed in part 2 when introducing the concept of sugarscape.

[tpike3]

I just pushed an update that focused on the scripting aspects.

Let me know if this works for you and makes the process more clear. The thought is this is our script not what the students see.

[tpike3]

@jackiekazil this is for part 3 -- part 2 will be an overview of sugarscape and the model

[tpike3]

I just pushed another update if you want to check out. The script is a loose guide, I rehearsed it and found as long as I have the general idea it is pretty easy to talk through it. Although my typing is pretty error ridden so I am likely not the the best one to record it.

[rht]

Using eval is dangerous pattern and overkill. I'd store the sugar and spice distribution data as 2 matrices (2 list-of-list). With the matrix, you can quickly do plt.imshow in the demo.

[tpike3]

Done ... it is much cleaner now .. thanks

[wang-boyu]

Would it make sense to include a section on how to integrate GIS data into ABM?

We can probably get some help from Andrew on class materials.

[rht]

I think the theme is to use various levels of Sugarscape models, building up features gradually. There is no Sugarscape model that uses GIS data (yet), so it's probably out of scope.

[tpike3]

@wang-boyu The goal is to build multiple modular pieces over time, start with the intro, then the advanced where we talk contributions and at some point build one for GIS, which could be sugarscape or something else.

Also keeping each block (intro, advanced etc) module as well will allow us to rerecord a section if we break something in the tutorial with an update.

[wang-boyu]

Sure! I completely understand.

Besides, Mesa-Geo is not yet ready for tutorials. Hopefully in the future we could have a section on ABM+GIS using Mesa & Mesa-Geo.