Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

Random Smoothing Regularization in Kernel Gradient Descent Learning

Authors: Liang Ding, Tianyang Hu, Jiahang Jiang, Donghao Li, Wenjia Wang, Yuan Yao

JMLR 2024 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct numerical experiments on simulated data to validate our theoretical results. In this section, we enhance our theoretical findings by experimentally validating the effectiveness of the random smoothing kernel with data augmentation and early stopping on synthetic data sets. We conduct numerical experiments on simulated data to validate our theoretical results. In this section, we present more details of numerical experiments conducted in Section 5.
Researcher Affiliation Collaboration Liang Ding EMAIL Fudan University Shanghai, China. Tianyang Hu EMAIL Huawei Noah s Ark Lab Shenzhen, China. Jiahang Jiang EMAIL The Hong Kong University of Science and Technology Hong Kong SAR, China. Donghao Li EMAIL The Hong Kong University of Science and Technology Hong Kong SAR, China. Wenjia Wang EMAIL The Hong Kong University of Science and Technology (Guangzhou) Guangzhou, China The Hong Kong University of Science and Technology Hong Kong SAR, China. Yuan Yao EMAIL The Hong Kong University of Science and Technology Hong Kong SAR, China.
Pseudocode No The paper describes mathematical derivations and methodological steps in paragraph form and equations, but it does not include any explicitly labeled pseudocode blocks or algorithm listings.
Open Source Code No The paper does not contain any explicit statements about releasing source code, nor does it provide links to code repositories.
Open Datasets Yes We conducted classification tasks on four real-world data sets: Iris (Fisher, 1988), Rice (Cammeo and Osmancik) (mis, 2019), Dry Bean (mis, 2020), and Raisin (C inar et al., 2023). DOI: https://doi.org/10.24432/C56C76. DOI: https://doi.org/10.24432/C5MW4Z. DOI: https://doi.org/10.24432/C50S4B. DOI: https://doi.org/10.24432/C5660T.
Dataset Splits Yes The simulated data are divided into the training set, validation set, and test set. The validation set is sampled as half the size of the training set, while the size of the test set is fixed at 500.
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments, such as GPU or CPU models.
Software Dependencies No The paper mentions training the neural network using stochastic gradient descent (SGD) but does not specify any software libraries (e.g., PyTorch, TensorFlow) or their version numbers.
Experiment Setup Yes We train the neural network using stochastic gradient descent (SGD) with momentum (0.9), small batch size (10), and learning rate β = 0.01. We choose a constant weight decay strength (10-4) to focus on the influence of random smoothing in cases with weight decay. The maximal step for SGD with early stopping is 100,000.