Interactive Machine Learning: Theory and Scale
Analysis
This dissertation addresses the challenges of acquiring labeled data and making decisions in machine learning, particularly in large-scale and high-stakes settings. It focuses on interactive machine learning, where the learner actively influences data collection and actions. The paper's significance lies in developing new algorithmic principles and establishing fundamental limits in active learning, sequential decision-making, and model selection, offering statistically optimal and computationally efficient algorithms. This work provides valuable guidance for deploying interactive learning methods in real-world scenarios.
Key Takeaways
- •Addresses challenges in acquiring labeled data and making decisions in machine learning.
- •Focuses on interactive machine learning where the learner actively influences data collection and actions.
- •Develops new algorithmic principles and establishes fundamental limits in active learning, sequential decision-making, and model selection.
- •Offers statistically optimal and computationally efficient algorithms.
- •Provides guidance for deploying interactive learning methods in real-world scenarios.
“The dissertation develops new algorithmic principles and establishes fundamental limits for interactive learning along three dimensions: active learning with noisy data and rich model classes, sequential decision making with large action spaces, and model selection under partial feedback.”