LOG: Active Model Adaptation for Label-Efficient OOD Generalization

Authors: Jie-Jing Shao, Lan-Zhe Guo, Xiao-wen Yang, Yu-Feng Li

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental A large number of experimental results confirm the promising performance of the new algorithm. In this section, we provide extensive results to evaluate LOG and compared methods for both benchmark simulation and a series of real-world tasks.
Researcher Affiliation Academia Jie-Jing Shao, Lan-Zhe Guo, Xiao-Wen Yang, Yu-Feng Li National Key Laboratory for Novel Software Technology Nanjing University, Nanjing 210023, China {shaojj, guolz, yangxw, liyf}@lamda.nju.edu.cn
Pseudocode No The paper includes a section titled 'Algorithm' (Section 4) but describes the process in prose rather than structured pseudocode or an algorithm block.
Open Source Code No The paper does not provide any explicit statements about releasing source code or a link to a code repository for the methodology described.
Open Datasets Yes In this task, we use a real-world dataset for car insurance prediction (Kaggle)... https://www.kaggle.com/anmolkumar/health-insurance-cross-sell-prediction. In this task, we use the Adult dataset... In this task, we use a real-world regression dataset of house sales prices from King County, USA... https://www.kaggle.com/c/house-prices-advanced-regression-techniques/data
Dataset Splits No The paper describes using a 'source labeled data DS' for training and evaluating on 'unlabeled data pool XT' (target distributions ET), but it does not specify traditional train/validation/test splits with percentages or sample counts for a validation set.
Hardware Specification No The paper does not provide specific details about the hardware used for running experiments, such as CPU or GPU models, memory, or cloud computing instance types.
Software Dependencies No The paper mentions using Logistic Regression and Linear Regression as base models and references methods like IGA [13] and [29], but it does not list any specific software libraries or their version numbers.
Experiment Setup Yes We generate 1000 samples from rs = 0.9 as source data DS, 5000 samples from 5 uniform environments ET with r [0.9, 0.7, 0.5, 0.3, 0.1] as unlabeled data pool XT (1000 samples for each r). We carry out the procedure 10 times and report the average results in Figure 2(a). Results on varying base distributions (under 10% labeling budgets).