Advanced Reinforcement Learning & Data Science for Social Impact
summary
Covers concrete methods and prototypes: uses inverse reinforcement learning and multi‑agent ensembles, builds a probabilistic model‑based RL pipeline to evaluate sequential HIV treatment decisions reduced to two stages from historical data, and simulates uncertainty rather than relying on a live simulator. Implements suitability mapping for wind farms on 5x5 km pixels, Google Maps scripts for road distance, and 12‑hour wind forecasts with deep learning. Also prototypes social‑media incident detection (semi‑supervised short‑text models), RFM customer segmentation and preliminary anomaly detection for electricity purchases; several projects are ongoing.
AI-written summary · about
chapters (13)
- 0:00Introduction and background
- 2:07Reinforcement learning research
- 7:28Inverse reinforcement learning explained
- 11:44Multiple reinforcement learning learners
- 13:48Evaluation in incomplete environments
- 17:44Data science for social impact
- 21:01Wind farm placement optimization
- 23:31Challenges in wind energy projects
- 29:13Electricity usage anomaly detection
- 33:32Public safety via social media
- 37:01Author identification in text
- 41:34Common challenges in social impact projects
- 45:07Closing remarks and future work
related talks
35mIndustrial Applications of Reinforcement Learning
32mData science vs real world: friends or foes
01h 51mAI and Social Entrepreneurship
46mData Discussion - AI for Good
29mReinforcement Learning Without a PhD: A Python Developer’s Journey
27mThe Journey from AI to Bedside: Translating Academic Research to Serving Patients Using Ray
30mReinforcement Learning - ChatGPT, Playing Games & More
31mDesigning game AI with reinforcement learning