A Difference Standardization Method for Mutual Transfer Learning

Authors: Haoqing Xu, Meng Wang, Beilun Wang

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

Reproducibility Variable Result LLM Response
Research Type Experimental We validate the method s performance using both synthetic and real-world data. Compared to previous methods, Diff S demonstrates a speed-up of approximately 3000 times that of similar methods and achieves the same accurate learnability structure estimation.
Researcher Affiliation Academia 1School of Computer Science and Engineering, Southeast University, Nanjing 210096, China 2Key Laboratory of Computer Network and Information Integration (Southeast University), Ministry of Education, China.
Pseudocode Yes Algorithm 1 Diff S ... Algorithm 2 Learnability Structure Recovering Ψ
Open Source Code No The paper states "We implement Diff S by Matlab, consistent with other baselines" but does not provide any specific link to source code or explicitly state that the code is publicly available.
Open Datasets Yes We apply Diff S on the NOAA n Clim Div Database1 (Vose et al., 2014) and Microarray Data (Wille et al., 2004). 1https://www.ncei.noaa.gov/access/ metadata/landing-page/bin/iso?id=gov.noaa. ncdc:C00005
Dataset Splits Yes All M domains are randomly splited into training set, validation set and test set, with the proportion of 7 : 1 : 2.
Hardware Specification Yes We implement Diff S by Matlab, consistent with other baselines, and conduct experiments on a Linux server with an Intel Xeon Bronze 3204 CPU with up to 12 threads and 32Gi B memory.
Software Dependencies No The paper states "We implement Diff S by Matlab", but it does not specify a version number for Matlab or any other software dependencies.
Experiment Setup Yes In the base case settings, we set the number of domains M = 50, the number of subgroups S = 5, the sample size n = 100, the dimension p = q = 10. ... We set ν = 0.001 for Diff S in all cases. For CD Fusion, the BIC tuning procedure optimizes the parameters for 5 times... For k-means, we vary k from 1 to M and choose the best k with the least BIC value.