Ideas List for Google Summer of Code

[tpike3]

Mesa Core:

• Improve/Rewrite Mesa Front End

Description: Update or rewrite the visualization stack. Currently Mesa uses a number of Javascript libraries for it's interactive charts and visualization, including Bootstrap (web interface), Chart.js (graphs) and D3 (network visualization). Outdated versions of these libraries are could be either updated or replaced. More chart types (like histograms, (stacked) bar charts or scatter charts) could be implemented to get quicker understand the agent behavior. Additional layouts of elements in the web interface could offer more flexible interfaces.

See issues: #1020, #1030, #1031, #1065, #1067, #1088, #1150 and (draft) PRs: #714, #846, #955, #1035, #1100, #1139, #1035

Expected Outcomes: An improved front end that allows for create user ease in creating a custom interface.

Skills Required: JavaScript, Python

Possible Mentors: @jackiekazil, @tpike3

Expected Size of Project: 350 Hours

Rating: medium to hard

• Improve performance

Find ways to optimize speed while balancing user ease. This can range from core optimizations such as better grid performance to extensions for HPCs/AWS etc. It can also include determining if there are additional optimizations that can be gained from advances in Python 3.10.

See issues: #167, #798, #882 and draft PRs: #815

Expected Outcomes: Core optimizations: Improved core mesa performance to run larger models and parameter sweeps. with minimal cost to user ease. Extensions: Supplementary libraries that allow users to more easily scale up to models on cloud platforms and HPCs

Skills Required: Core Optimizations: Python; Extensions: Python, C, C++

Possible Mentors: @tpike3 , @jackiekazil, @EwoutH

Expected Size of Project: Core Optimizations : 175 Hours; Extensions: 350 Hours

Rating: Core Optimizations : medium ; Extensions: hard

• Implement native support for multiple agent types.

Currently Mesa only supports having one agent type properly, and it's schedulers, spaces (grids) data collection and visualization are not able to handle more than one without modifications. It should be comparable to breeds in NetLogo.

See draft-PR #1142 for a proof of concept, #1145 as some work on the datacollection and visualisation and #1162 as an implemenation of a scheduler.

Expected Outcomes: Users have improved agent scheduler to allow for a variety of agent types and variation in scheduling order. Updates ensure data collector and front end compatibility

Skills Required: Python

Possible Mentors: @EwoutH, @tpike3

Expected Size of Project: 175 Hours

Rating: Medium

Other Areas for Improvement:

Mesa Ecosystem:

• Improve and expand Corvine/Mesa-Geo
  --Description: Mesa-Geo was built by an outstanding Mesa volunteer who based on changing circumstances could no longer support the project. As a critical extension we would like to reinvigorate and get mesa-geo as part of the Mesa ecosystem.
  -- Expected Outcomes: Move Geo-Mesa into new Mesa extension repository, ensure compatible with Mesa updates and address outstanding issues/ bugs
  -- Skills Required: GIS, Python
  -- Possible Mentors: @tpike3, @jackiekazil
  -- Expected Size of Project: 175 Hours
  -- Rating: Medium

• Improve and Expand Mesa Data
  --Description: Having transparent, accessible and extensible data pipelines we feel is critical to building a robust ABM ecosystem that has the ease of use of existing ML libraries. This is an initial attempt to grow such an ecosystem.
  -- Expected Outcomes: Synthetic Population: Update population smoothing to use a convolution approach; update sub-population selection to use organic Bokeh library selection tools. Build a Data Pipeline: Identify a desired data pipeline for an ABM and build it using and improving upon the existing format.
  -- Skills Required: Python
  -- Possible Mentors: @tpike3
  -- Expected Size of Project: Synthetic Population:175 Hours ; Build a Data Pipeline: 350 Hours
  -- Rating: Easy to Medium

• Improve and Expand Mesa Behaviors
  --Description: We assess another critical part of the ecosystem is behaviors that users can easily import into their models. This is an initial effort to build out a library of such behaviors
  -- Expected Outcomes: Multi-level Mesa: Re-envision multi-level Mesa using hypergraphs instead of standard networks . Build a behavior: Identify a desired agent behavior for an ABM and build it using.
  -- Skills Required: Python
  -- Possible Mentors: @tpike3
  -- Expected Size of Project: Multi-Level Mesa:350 Hours ; Build a Behavior: 350 Hours
  -- Rating: Hard

Mesa Education:

• Help build the Mesa Complexity Explorer Tutorial
  -- Expected Outcomes: Completed a Mesa tutorial that runs through building a model and contributing to Mesa (in this case a model example)
  -- Skills Required: Python; Skills Desired: MOOC education and platforms
  -- Possible Mentors: @tpike3, @jackiekazil , @EwoutH
  -- Expected Size of Project: 175 Hours
  -- Rating: Easy to Medium