A Probabilistic Covariate Shift Assumption for Domain Adaptation

Authors: Tameem Adel, Alexander Wong

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

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our proposed algorithm on a benchmark sentiment analysis (and domain adaptation) dataset, where state-of-the-art adaptation results are achieved.
Researcher Affiliation Academia Tameem Adel University of Waterloo thesham@uwaterloo.ca Alexander Wong University of Waterloo a28wong@uwaterloo.ca
Pseudocode No The paper describes the algorithm and its steps using text and mathematical formulations but does not include any explicitly labeled pseudocode or algorithm blocks.
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 To evaluate the performance of the proposed algorithm, we run our experiments on the Amazon reviews dataset (Blitzer, Dredze, and Pereira 2007).
Dataset Splits No The paper mentions 'Parameter values are determined by 10-fold cross-validation (K = 7 for Prob CS),' which is a validation strategy. However, it does not specify explicit fixed training, validation, and test dataset splits by percentages, absolute counts, or specific predefined splits for the main experimental evaluation.
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., CPU models, GPU models, memory, or cloud computing specifications) used to run the experiments.
Software Dependencies No The paper mentions using a 'linear SVM learner' and 'sequential minimal optimization (SMO)', and states that for comparison, 'CODA, SDA and m SDA, we used the implementation provided by the authors.' However, it does not provide specific version numbers for its own software dependencies, such as Python libraries or frameworks.
Experiment Setup Yes We train a linear SVM on S, the labeled 2000 instances (reviews) of a source domain. [...] It is a linear SVM learner with C = 1. [...] we select the 5,000 most frequent features (vocabulary terms of unigrams and bigrams). [...] For each instance xi in T, set the corresponding υ(ψ) according to fsvm(xi). Note that for xi not within the SVM margin, υ(ψ) = 0, thus Pr(l S(xi) = l T (xi)) = 0 and they retain their source labels. [...] Parameter values are determined by 10-fold cross-validation (K = 7 for Prob CS).