Agenda for 24 Jan

[mikeizbicki]

1. Scheduling a recurring meeting. (8:30am 9:30am Tues)

2. Background info about the project.

   Project repo: https://github.com/SelfExplainML/PiML-Toolbox

3. Assign preliminary roles.

   Need to contact: Nengfeng.Zhou@wellsfargo.com, take note post to github

[agallagher23]

Week One Tasks:

• Set up weekly recurring meeting with contact
• Go through a low-code library example (California Housing Data and one #5-8), note confusions & ideas to improve
• Explore links below, save links you find really helpful

Potential directions

1. Add a low-code example with a public dataset
2. Add a high-code example (translating low-code to high-code)
3. Youtube/blog post tutorial explaining application high goal

[mikeizbicki]

Some interesting links:

1. https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing
2. https://medium.com/hackernoon/more-than-a-million-pro-repeal-net-neutrality-comments-were-likely-faked-e9f0e3ed36a6
3. https://www.nytimes.com/2012/02/19/magazine/shopping-habits.html?pagewanted=1&_r=1&hp
4. https://twitter.com/jackyalcine/status/615329515909156865
5. https://theblog.okcupid.com/exactly-what-to-say-in-a-first-message-2bf680806c72

[agallagher23]

1/30 WF: Nengfeng Zhou, Rosh

Meeting notes

• Nengfeng took us through the PiML GitHut page
• Generally our goal is to…
  • Understand framework PiML team built
    • How are they different from other black box models? Can this model achieve similar performance?
  • Understand model explainability and variable importance
  • Forward questions we have about documentation on the page to share with the PiML team
  • Provide a case study/use experience to uncover where the model is not doing well enough
    • Use data related to financial industry

Next steps

• Send WF our GitHub page
• Research general concepts
  • Black box models
  • Model explainability
  • Variable Importance
• Explore resources WF sends to us

[everettbu]

What we can help with:

• Adding code is not easy
• Needs help improving user experience
• Feedback on how to make it more user friendly
• Improve internal model
  • What situations can it be improved
  • Identify weaknesses of the models