Covariate-Shift Generalization via Random Sample Weighting
Authors: Yue He, Xinwei Shen, Renzhe Xu, Tong Zhang, Yong Jiang, Wenchao Zou, Peng Cui
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on both simulated and real-world datasets clearly validate the effectiveness of our method. and 4 Experiment In this section, we evaluate the effectiveness of proposed method RSW towards covariate-shift generalization, in comparison with the benchmark models. We carry out extensive experiments on both synthetic data and real-world datasets where the distribution shifts exist. |
| Researcher Affiliation | Collaboration | Yue He1, Xinwei Shen2, Renzhe Xu1, Tong Zhang3, Yong Jiang1, Wenchao Zou4, Peng Cui1* 1Tsinghua University 2ETH Z urich 3The Hong Kong University of Science and Technology 4Siemens |
| Pseudocode | Yes | Algorithm 1: Covariate-shift Generalization via Random Sample Weighting (RSW) |
| Open Source Code | No | The paper does not contain an explicit statement about releasing the source code or provide a link to a code repository for the described methodology. |
| Open Datasets | Yes | In this experiment, we use a real-world regression dataset1 (Kaggle) of house sales prices from King County, USA. Specifically, we split the dataset into 6 periods (each covers a time span of two decades) between 1900 2015 according to the built year of house. 1https://www.kaggle.com/datasets/harlfoxem/housesalesprediction and In this experiment, we use the Adult dataset2 (Kohavi 1996), of which the task is to predict whether the personal income exceeds 50K/yr based on census data. 2https://archive.ics.uci.edu/ml/datasets/adult and In this experiment, we use the CS-Colored MNIST dataset (Ahuja et al. 2021) that simulates the covariate-shift in image data. |
| Dataset Splits | Yes | Then we optimize the models using the 1000 training samples collected from a single environment where r = 2/3. and we train all the methods on the first period ([1900, 1919]) and test if they can predict the transaction price on the other periods respectively. and We split the dataset into 10 environments according to the combination of attributes race and sex. Like in House Price Prediction, we train all the methods on the first environment (White, Female) and test them on the other environments respectively. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., exact GPU/CPU models, processor types with speeds, or memory amounts) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers (e.g., library or solver names with version numbers like Python 3.8, PyTorch 1.9) needed to replicate the experiment. |
| Experiment Setup | Yes | Simulation 1: p = 10, |Vb| = 2, Scale = 6, Simulation 2: p = 10, |Vb| = 2, r = 2, Simulation 3: |Vb|/|V | = 0.1, r = 0.2, Scale = 6, A 3-Layer MLP is taken as prediction model for image |