Task Utility
Task utility, in the context of machine learning and AI, focuses on evaluating the usefulness and effectiveness of algorithms and models in achieving specific tasks, often considering trade-offs with other factors like privacy, fairness, and computational efficiency. Current research emphasizes developing robust benchmarking methods and algorithms to measure and improve utility across diverse applications, including data synthesis, optimization problems, and human-AI collaboration, employing techniques like generative adversarial networks, spectral graph learning, and temporal point processes. Understanding and maximizing task utility is crucial for responsible AI development, ensuring that models not only perform well but also deliver practical value and align with ethical considerations in real-world deployments.
Papers
Post-processing Private Synthetic Data for Improving Utility on Selected Measures
Hao Wang, Shivchander Sudalairaj, John Henning, Kristjan Greenewald, Akash Srivastava
Theoretically Principled Federated Learning for Balancing Privacy and Utility
Xiaojin Zhang, Wenjie Li, Kai Chen, Shutao Xia, Qiang Yang