Out-Of-Domain Unlabeled Data Improves Generalization

Authors: seyed amir hossein saberi, Amir Najafi, Alireza Heidari, Mohammad Hosein Movasaghinia, Abolfazl Motahari, Babak Khalaj

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental From a theoretical standpoint, we apply our framework on the classification problem of a mixture of two Gaussians in Rd... We validate our claims through experiments conducted on a variety of synthetic and real-world datasets.
Researcher Affiliation Academia Department of Electrical Engineering, Department of Computer Engineering, Sharif Center for Information Systems and Data Science, Sharif Institute for Convergence Science & Technology, Sharif University of Technology, Tehran, Iran
Pseudocode Yes Algorithm 1 Finding the adversarial perturbed input for original input data based on gradient ascent
Open Source Code No The paper does not provide a direct link to a source-code repository nor explicitly states that the code for their method is being released.
Open Datasets Yes NCT-CRC-HE-100K consists of 100,000 histopathology images of colon tissue (Katherm et al., 2018).
Dataset Splits No Finally, we select a combination of hyper-parameters that achieved the highest accuracy on a validation dataset, and we report the accuracy of our model, using these hyper-parameters, on the test samples.
Hardware Specification No The paper does not explicitly describe the specific hardware (e.g., GPU models, CPU models, or memory) used to run its experiments.
Software Dependencies Yes The codes are written using the Python programming language and the Pytorch 2.0 machine learning framework.
Experiment Setup Yes A random search process has been performed to find the optimum γ, γ , λ, and weight-decay. Finally, we select a combination of hyper-parameters that achieved the highest accuracy on a validation dataset, and we report the accuracy of our model, using these hyper-parameters, on the test samples.