Stable Prediction with Model Misspecification and Agnostic Distribution Shift

Authors: Kun Kuang, Ruoxuan Xiong, Peng Cui, Susan Athey, Bo Li4485-4492

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments clearly demonstrate that our DWR algorithm can significantly improve the accuracy of parameter estimation and stability of prediction with model misspecification and agnostic distribution shift. Experiments In this section, we check the performance of our algorithm with experiments on both synthetic and real-world datasets.
Researcher Affiliation Academia 1Zhejiang University 2Tsinghua University 3Stanford University
Pseudocode Yes Algorithm 1 Decorrelated Weighted Regression algorithm
Open Source Code Yes The online appendix and supplementary materials are available at http://kunkuang.github.io or https://www.dropbox.com/s/1q0brkc2bnehhfo/paperaaai20-Supplementary.
Open Datasets Yes We collected air pollutant data and meteorological data from the U.S. EPA s Air Quality System (AQS) database,4 which has been widely used for model evaluation (Yahya et al. 2017; Zhu et al. 2018). https://www.epa.gov/outdoor-air-quality-data
Dataset Splits Yes we trained all models with data from State 1, validated with data from States 1 to 4, finally tested them on all 10 States.
Hardware Specification No The paper does not explicitly describe the hardware used for experiments. It mentions 'a machine learning problem' but no specific GPU/CPU models or cloud instance types.
Software Dependencies No The paper does not provide specific software dependencies with version numbers.
Experiment Setup Yes In our experiments, we tuned these parameters with cross validation by grid searching, and each parameter is uniformly varied from {0.01, 0.1, 1, 10, 100}. Initialize parameters W (0) and β(0)