Connecting Sphere Manifolds Hierarchically for Regularization

Authors: Damien Scieur, Youngsung Kim

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We experimented the proposed method using five publicly available datasets, namely CIFAR100 (Krizhevsky, 2009), Caltech-UCSD Birds 200 (CUB200) (Welinder et al., 2010), Stanford-Cars (Cars) (Krause et al., 2013), Stanford-dogs (Dogs) (Khosla et al., 2011), and Tiny-Image Net (Tiny Im Net) (Deng et al., 2009).
Researcher Affiliation Industry 1Samsung SAIT AI Lab, Montreal 2Samsung Advanced Institute of Technology (SAIT).
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any statement or link indicating that the source code for the methodology is openly available.
Open Datasets Yes We experimented the proposed method using five publicly available datasets, namely CIFAR100 (Krizhevsky, 2009), Caltech-UCSD Birds 200 (CUB200) (Welinder et al., 2010), Stanford-Cars (Cars) (Krause et al., 2013), Stanford-dogs (Dogs) (Khosla et al., 2011), and Tiny-Image Net (Tiny Im Net) (Deng et al., 2009).
Dataset Splits No The paper mentions training and testing but does not provide specific dataset split information (exact percentages, sample counts, or detailed splitting methodology for training, validation, and test sets).
Hardware Specification Yes All tests are conducted using NVIDIA Tesla V100 GPU with the same random seed.
Software Dependencies No The paper mentions using 'Geoopt optimizer (Kochurov et al., 2020)' but does not provide specific version numbers for software components or libraries.
Experiment Setup Yes We used the stochastic gradient descent (SGD) over 300 epochs, with a mini-batch of 64 and a momentum parameter of 0.9 for training. The learning rate schedule is the same for all experiments, starting at 0.1, then decaying by a factor of 10 after 150, then 225 epochs.