Heterogeneous Transfer Learning with Weighted Instance-Correspondence Data
Authors: Yuwei He, Xiaoming Jin, Guiguang Ding, Yuchen Guo, Jungong Han, Jiyong Zhang, Sicheng Zhao4099-4106
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct extensive experiments on He TL datasets and the state-of-the-art results verify the effectiveness of TLWC. |
| Researcher Affiliation | Academia | Yuwei He,1 Xiaoming Jin,1 Guiguang Ding,1 Yuchen Guo,1,2 Jungong Han,3 Jiyong Zhang,4 Sicheng Zhao5 1Beijing National Research Center for Information Science and Technology (BNRist) 1School of Software, 2Department of Automation, Tsinghua University, Beijing 100084, China 3WMG Data Science, University of Warwick, Coventry, UK 4School of Automation, Hangzhou Dianzi University, China 5Department of EECS, University of California, Berkeley, USA |
| Pseudocode | Yes | Algorithm 1 Transfer Learning with Weighted Correspondence |
| Open Source Code | No | The paper does not contain any explicit statements or links indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | We adopt the widely used 9 He TL tasks of this dataset: EFB, EFD, EFM, EGB, EGD, EGM, EJB, EJD, EJM. EFB means, for example, taking book reviews in English as the source domain and those in French as the target domain. the documents are represented with TF-IDF and 2000 most frequent words are selected. Multilingual Reuters Collection (MRC) is a news dataset with five languages (English (EN), French (FR), German (GE), Italian(IT) and Spanish(SP)), where each article is represented by TF-IDF. NUS-WIDE (Chua et al. 2009) contains 269,648 images from Flickr and their corresponding text-tag. |
| Dataset Splits | Yes | The given data are: sufficient labeled data {XS,l, y S,l} = {(x S,l i , y S i )}n S,l i=1 and unlabeled data {XS,u} = {x S,u i }n S,u i=1 in the source domain; a set of labeled data {XT,l, y T,l} = {(x T,l i , y T i )}n T,l i=1 and unlabeled data {XT,u} = {x T,u i }n T,u i=1 in the target domain, where n S,l n T,l; a set of unlabeled IC data {XS,c, XT,c} = {(x S,c i , x T,c i )}nc i=1 across the two domains. Table 1: The data volume for experiments. |XS,c| |XS,l| |XS,u| |XT,c| |XT,l| |XT,u| Test Amazon 2000 2000 9000 2000 100 9000 1900 MRC 500 3000 0 300 30 0 3000 NUS-WIDE 500 5000 0 500 100 0 1000 |
| Hardware Specification | No | The paper does not specify any hardware details such as GPU/CPU models or types of computing resources used for the experiments. |
| Software Dependencies | No | The paper mentions methods and algorithms but does not provide specific software names with version numbers for reproducibility (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | Table 2: The parameter settings for experiments. K λ η α β γ τ Amazon 3 0.7 1 0.01 105 2 0.01 MRC 3 0.7 1 0.01 105 2 0.01 NUS-WIDE 3 0.7 1 0.01 105 2 10 5 |