Supervised Representation Learning: Transfer Learning with Deep Autoencoders

Authors: Fuzhen Zhuang, Xiaohu Cheng, Ping Luo, Sinno Jialin Pan, Qing He

IJCAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we conduct extensive experiments on three real-world image data sets to show the effectiveness of the proposed framework. Two of the three datasets are on binary classification, and the rest one is on multi-class classification. All the results of these three data sets are shown in Figure 2 and Table 3.
Researcher Affiliation Academia 1Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China. {zhuangfz, heq}@ics.ict.ac.cn, luop@ict.ac.cn 2University of Chinese Academy of Sciences, Beijing, China. chengxh@ics.ict.ac.cn 3Nanyang Technological University, Singapore 639798. sinnopan@ntu.edu.sg
Pseudocode Yes Algorithm 1 Transfer Learning with Deep Autoencoders (TLDA)
Open Source Code No The paper mentions using 'authors source code3' for the baseline method m SDA (footnote 3 points to http://www.cse.wustl.edu/ mchen/), but it does not provide a link or statement about the availability of the source code for their proposed TLDA method.
Open Datasets Yes Image Net Data Set1 contains five domains, i.e., D1 (ambulance+scooter), D2 (taxi+scooter), D3 (jeep+scooter), D4 (minivan+scooter) and D5 (passenger car+scooter). Data from different domains come from different categories, e.g., taxi from D2 and jeep from D3, therefore this dataset is 1http://www.image-net.org/download-features proper for transfer learning study. Corel Data Set2 [Zhuang et al., 2010] includs two different top categories, flower and traffic. 2http://archive.ics.uci.edu/ml/datasets/Corel+Image+Features. Leaves Data Set [Mallah and Orwell, 2013] includes 100 plant species that are divided into 32 different genera, and each specie has 16 instances.
Dataset Splits No The paper describes how classification problems are constructed from datasets (e.g., 'we construct 20 (P 2 5 ) transfer learning classification problems'), but it does not provide specific percentages, absolute sample counts, or explicit cross-validation details for training, validation, or test splits for its experiments.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU models, CPU types, memory) used for running the experiments.
Software Dependencies No The paper mentions using Logistic Regression, TCA, and m SDA, and refers to using the source code for m SDA (footnote 3: http://www.cse.wustl.edu/ mchen/), but it does not specify version numbers for any software, libraries, or programming languages used.
Experiment Setup Yes After some preliminary experiments, we set α = 0.5, β = 0.5, γ = 0.00001 and k = 10 for the Image Net and Corel datasets, while β = 0.05, k = 5 and γ = 0.0001 for the Leaves dataset. For m SDA, we use the authors source code3 and adopt the default parameters as reported in [Chen et al., 2012]. For TCA, the number of latent dimensions is carefully tuned, e.g., for the Corel dataset, the number is sampled from [10, 80] with interval 10, and its best results are reported.