Overcoming Concept Shift in Domain-Aware Settings through Consolidated Internal Distributions

Authors: Mohammad Rostami, Aram Galstyan

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct experiments on five benchmarks and observe that our algorithm compares favorably against SOTA UDA methods.Empirical Evaluation: Since sequential model adaptation is not a well-explored problem, we follow the UDA literature for evaluation due to the topic proximity.
Researcher Affiliation Academia Information Sciences Institute, University of Southern California {mrostami, galstyan}@isi.edu
Pseudocode Yes Algorithm 1: SDAUP (λ, ITR)
Open Source Code Yes Our code is provided at https://github.com/rostami-m/SDAUP.
Open Datasets Yes We validate our method on five standard UDA benchmarks and adapted them for sequential task learning: Digit recognition tasks, Office-31 Dataset, Image CLEF-DA Dataset, Office-Caltech Dataset, and Vis DA-2017. Details about these datasets are included in the Appendix.
Dataset Splits No The paper mentions using source and target datasets but does not explicitly provide specific numerical training, validation, and test split percentages or counts for any of the datasets.
Hardware Specification No The paper does not provide specific hardware details such as GPU or CPU models used for running the experiments.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers, such as programming languages, libraries, or frameworks (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes A point of strength for our algorithm is that there are only two major algorithm-specific hyper-parameters and tuning them is not challenging. We set τ = 0.99 and λ = 10^-3.