Reproduce “Deep Hashing Network for Unsupervised Domain Adaptation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Hemanth Venkateswara, Jose Eusebio, Shayok Chakraborty, Sethuraman Panchanathan

Researchers at Arizona State University developed the Domain Adaptive Hashing (DAH) network, a deep learning framework that learns representative binary hash codes to simultaneously address unsupervised domain adaptation and provide efficient data representations. The DAH model achieved an average object recognition accuracy of 45.54% on the newly introduced Office-Home dataset and consistently outperformed existing unsupervised hashing methods in Mean Average Precision (mAP).

Key insights

  • The DAH framework is the first to integrate deep feature learning with hash code generation specifically for unsupervised domain adaptation, unifying category assignment and efficient data representation.
  • Minimizing entropy on unlabeled target data effectively guides the network to learn discriminative hash values, enhancing target sample alignment with source categories.
  • The proposed loss function and network architecture significantly reduce domain discrepancy in the feature space, leading to more transferable and discriminative representations.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.