Active Prediction in Dynamical Systems

Peer Neighborhood Mechanisms: A Framework for Mechanism Generalization

“…One attempt is the Logarithmic Peer Truth Serum (LPTS), which does so by assuming a locality structure among the Agents so that more localized Peers have more similar statistical properties (Radanovic and Faltings 2015), and the Personalized Peer Truth Serum, which extends the LPTS for subjective private data (Goel and Faltings 2020). Finally, the work of (Chen, Shen, and Zheng 2020) and (Kong and Schoenebeck 2019) consider mechanisms which reward Agents according to the mutual information between reports, but assume that the distribution can be parameterized as an element of a known family of distributions. We see that in all cases, the ability to extend Peer Consistency concepts to arbitrary distributions relies on a priori assumptions about the structure of the distributions or the Agent properties.…”

HAPI Explorer: Comprehension, Discovery, and Explanation on History of ML APIs

“…The dynamics of API changes have been studied for a while as the MLaaS market grows bigger (Chen, Zaharia, and Zou 2020). They mainly focus on the prices of calling APIs (Chen, Koutris, and Kumar 2019) and prediction results analysis of the APIs (Hosseini, Xiao, and Poovendran 2017).…”

“…Machine learning (ML) prediction APIs offered by Google, Microsoft, Amazon, and many other providers have been continuously adopted in a plethora of applications, such as visual object detection, natural language comprehension, and speech recognition (Chen, Zaharia, and Zou 2020;Tramèr et al 2016). Reliable ML deployments require systematical understanding and comparison of different APIs in varying aspects, including overall accuracy (Koenecke et al 2020), inherent biases (Buolamwini and Gebru 2018), and prediction consistency over time (Chen, Zaharia, and Zou 2021).…”