In search of robust measures of generalization

Authors: Gintare Karolina Dziugaite, Alexandre Drouin, Brady Neal, Nitarshan Rajkumar, Ethan Caballero, Linbo Wang, Ioannis Mitliagkas, Daniel M. Roy

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We collect data from thousands of experiments on CIFAR-10 and SVHN, with various values for width, depth, learning rate, and training set size.
Researcher Affiliation Collaboration 1Element AI, 2Mila, 3Université de Montréal, 4University of Toronto, 5Vector Institute
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide an explicit statement about releasing its source code or a link to a repository for the methodology described.
Open Datasets Yes We collect data from thousands of experiments on CIFAR-10 and SVHN, with various values for width, depth, learning rate, and training set size.
Dataset Splits No The paper mentions 'training data' and 'held-out data' but does not explicitly provide specific train/validation/test dataset split percentages, sample counts, or a detailed splitting methodology for reproducibility.
Hardware Specification No The paper mentions 'massive computing resources' but does not provide specific hardware details such as GPU/CPU models, processor types, or memory specifications used for running experiments.
Software Dependencies No The paper refers to software such as PyTorch, NumPy, Matplotlib, and Scikit-learn in its references, but it does not specify version numbers for these or any other software dependencies needed to replicate the experiments.
Experiment Setup Yes In our experiments, variable assignments (!) are pairs (H, σ) of hyperparameter settings and random seeds, respectively. The hyperparameters are: learning rate, neural network width and depth; dataset (CIFAR-10 or SVHN), and training set size. (See Appendix C for ranges.)