Applied Machine Learning for Social Good Workshop
Thursday,October 29, 8am-Noon PT
- Rakshit Agrawal, Camio Inc.
Organizer of Applied ML for Social Good Tutorial at GHTC 2019
- Charles Delahunt, Intellectual Ventures Lab / Global Good, Bellevue, WA
University of Washington, Applied Math Dept
This session will focus on topics around the usage of machine learning (ML) methods in order to assist social good applications. We will discuss a framework centered around identifying the important societal challenges where problems can be defined for ML capabilities. We then describe the process to develop machine learning models and deploy them in the real world.
The workshop is 4 hours:
- Introduction to the concepts of developing ML solutions for social good applications.
- Several short invited talks (~25 minutes) on key topics, connected together with contextual material, along with Q/A.
- Closing, further resources, questions and discussion as wished.
- First we will introduce the framework of Identification, Definition, Solution and Deployment.
- Then we will discuss the identification using UN Global Goals.
- This will then be used in definition using targets from Global Goals.
- Then we will teach a Machine Learning pipeline (Gather data, Extract features, Develop model, Train and optimize, Evaluate).
- This pipeline will then be illustrated during the talks with specific examples.
IVL GG = Intellectual Ventures Lab, Global Good. Bellevue, WA.
IST = Intelligent Systems Group (at IVL GG)
LMIC = Low-Middle Income Countries
US = Ultrasound
1. Commercialization: Craig Nakagawa (IVL GG, Partner Development).
a. Receptiveness to AI in LMIC – much higher willingness to adopt, due to pressing needs and lack of resistance from existing players.
b. Importance of engagement by local groups and partners for successful deployments. Example: health care clusters in LMIC.
c. Partnering with the private sector for large-scale deployment
2. Identifying and defining projects: Mark Newell (IVL GG, Global Development Technologies Portfolio)
a. Working with field partners to identify target problems
b. Researching appropriate product performance specs to ensure impact
c. Data collection for project development
3. Defining metrics to guide ML development: Charles Delahunt (IVL GG, ISG; and U. Washington Applied Math)
a. Limits of standard ML metrics (AUC, object accuracy)
b. Importance of the product performance specs (as opposed to generic improvement in accuracy)
c. Tailoring/defining the metrics of model outputs to the particular task
d. Examples: Malaria detection in blood film images; Grain moisture prediction.
4. Case study: Lung Ultrasound: Rachel Millin (IVL GG, ISG)
a. Overview of the IVL lung ultrasound project, including
c. commercial partners and their different product goals;
d. data, data collection, and annotation;
e. metrics for model evaluation;
f. ML models;
g. Some results.
5. [Tentative] Introduction to Deep Neural Nets: Sourabh Kulhare (IVL GG, ISG)
b. Some system details (data, annotation, architectures, etc)
d. Tips/observations/pitfalls of set-up and training
At the end of the session, we expect that the participants will develop a systematic flow of thinking about solutions using machine learning for tackling societal challenges, as well as familiarity with some key principles, grounded in several concrete examples. In addition, participants will have a clearer idea of the landscape, and of necessary considerations for effective deployments.
This tutorial will help build a framework which can be used to build different solutions. With the help of discussions in the session, we expect to build a strong community in this domain, which focuses on deployable solutions for real societal challenges, and help achieve some progress on the United Nations Global Goals.
- Registration Open for GHTC 2020 August 12, 2020
- Preliminary Speaker Announcement June 3, 2020
- Toshio Fukuda June 3, 2020
- Kathy Land June 3, 2020
- Melissa Sassi June 3, 2020
Visit us on:
Your logo could be here