Dr. Pradeep Ravikumar addresses the pressing challenges of deploying machine learning models in high-stakes environments, particularly the need for robustness against adversarial inputs that can drastically alter predictions. Highlighting a novel strategy, he explores the use of an ensemble of neural networks designed to defend against such threats while ensuring high performance on the least favorable samples. This presentation bridges crucial areas such as algorithmic fairness, class imbalance, and risk-sensitive decision-making, making it invaluable for professionals and enthusiasts in machine learning, AI, data science, and more.
#MachineLearning #ArtificialIntelligence #DataScience #RobustAI #AdversarialAI #DeepLearning #AIResearch #DataEngineering #NLP #DataVisualization #TechTalk #AIInnovation
Timecodes:
0:00 - Intro
0:55 - Distribution Shift
4:31 - Unknown Sub-populations, Tail Risk
6:38 - CVAR&DRO
14:28 - Deterministic to Randomized
19:48 - Solving Zero Sum Games
35:25 - DORO: DIstributional Outlier Robust Optimization
42:15 - Summary
Want to stay ahead on AI? Then subscribe to ODSC today 🚀🧠- https://www.youtube.com/c/OpenDataSci...
You can also follow ODSC on:
LinkedIn - / open-data-science
Twitter - / _odsc
Facebook - / opendatasci
Medium - / odsc