Learning and Optimization of Static and Dynamic Appointments
summary
Proposes a model-free, data-driven method that learns appointment times from observed service samples using stochastic gradient with infinitesimal perturbation analysis, yielding a convex static-scheduling problem that converges to the optimal schedule. Extends to real-time warning-time policies parameterized by thresholds X, which cut cascaded delays by notifying customers adaptively; dynamic problems lose global convexity so the algorithm uses multiple restarts and handles practical instances up to about ten surgeries, requiring realtime signaling from staff.
AI-written summary · about
chapters (9)
- 0:00Hospital scheduling problem
- 1:49Defining appointment models
- 3:41Theoretical challenges
- 5:18Model-free approach
- 7:34Linear programming baseline
- 12:03Stochastic gradient descent
- 16:44Real-time scheduling
- 21:56Results and practicality
- 23:42Implementation challenges
related talks
31mAvoiding Cascading Failures for Time-of-Use Pricing
40mApproximation Algorithms for Optimization under Uncertainty
35mOnline Optimization and Learning Under Long-Term Convex Constraints and Objectives
40mA Bayesian Approach to Online Resource Allocation
49mScheduling with Time-Evolving Uncertainty for Content Review Prioritization in Social Media
45mKeynote: Seeking Order amidst Chaos and Uncertainty
16mIntelligent Resource Scheduling for Co-located Latency-critical Services: A Multi-Model
28mConfounding-Robust Policy Evaluation In Infinite-Horizon Reinforcement Learning