Learning Semantic Representations for Unsupervised Domain Adaptation

Authors: Shaoan Xie, Zibin Zheng, Liang Chen, Chuan Chen

ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments testify that our model yields state of the art results on standard datasets.
Researcher Affiliation Academia 1School of Data and Computer Science, Sun Yat-sen University, Guangzhou, China 2National Engineering Research Center of Digital Life, Sun Yat-sen University, Guangzhou, China.
Pseudocode Yes Algorithm 1 Moving semantic transfer loss computation in iteration t in our model.
Open Source Code Yes Codes are available at https://github.com/Mid-Push/Moving-Semantic-Transfer-Network.
Open Datasets Yes Office-31 (Saenko et al., 2010) is a standard dataset used for domain adaptation. MNIST (Le Cun et al., 1998), USPS and SVNH (Netzer et al., 2011).
Dataset Splits No The paper mentions applying “reverse validation” for hyperparameter tuning but does not explicitly provide percentages or counts for training/validation/test dataset splits for the experiments.
Hardware Specification No The paper mentions the network architectures used (Alex Net, CNN) but does not specify any hardware details such as CPU/GPU models, memory, or cloud resources used for experiments.
Software Dependencies No The paper discusses optimization methods (Stochastic gradient descent) but does not specify software dependencies with version numbers (e.g., Python version, PyTorch/TensorFlow version, CUDA version).
Experiment Setup Yes We set θ=0.7 in all our experiments. For the weight balance parameter, we set λ = 2 1+exp( γ.p) 1, where γ is set to 10 and p is training progress changing from 0 to 1. ... Stochastic gradient descent with 0.9 momentum is used. The learning rate is annealed by µp = µ0 (1+α.p)β , where µ0=0.01, α=10 and β=0.75... We set the batch size to 128 for each domain.