Towards closing the gap between the theory and practice of SVRG

Authors: Othmane Sebbouh, Nidham Gazagnadou, Samy Jelassi, Francis Bach, Robert Gower

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We performed a series of experiments on data sets from LIBSVM [5] and the UCI repository [3], to validate our theoretical findings. We tested l2 regularized logistic regression on ijcnn1 and real-sim, and ridge regression on slice and Year Prediction MSD. We used two choices for the regularizer: λ = 10 1 and λ = 10 3. All of our code is implemented in Julia 1.0. Due to lack of space, most figures have been relegated to Section G in the supplementary material.
Researcher Affiliation Academia Othmane Sebbouh LTCI, T el ecom Paris Institut Polytechnique de Paris othmane.sebbouh@gmail.com Nidham Gazagnadou LTCI, T el ecom Paris Institut Polytechnique de Paris nidham.gazagnadou@telecom-paris.fr Samy Jelassi ORFE Department Princeton University sjelassi@princeton.edu Francis Bach INRIA Ecole Normale Sup erieure PSL Research University francis.bach@inria.fr Robert M. Gower LTCI, T el ecom Paris Institut Polytechnique de Paris robert.gower@telecom-paris.fr
Pseudocode Yes Algorithm 1 Free-SVRG Algorithm 2 L-SVRG-D
Open Source Code No The information is insufficient. The paper states 'All of our code is implemented in Julia 1.0.' but does not provide a link or explicit statement of open-source availability for the described methodology.
Open Datasets Yes We performed a series of experiments on data sets from LIBSVM [5] and the UCI repository [3]
Dataset Splits No The information is insufficient. The paper does not explicitly provide details about training, validation, or test dataset splits. It mentions 'training problems' and 'training sets' in a general context but not specific data partitioning for experiments.
Hardware Specification No The information is insufficient. The paper does not provide any specific hardware details (e.g., CPU/GPU models, memory, or cloud instance types) used for running the experiments.
Software Dependencies No The information is insufficient. The paper only states 'All of our code is implemented in Julia 1.0.' without specifying any versioned libraries or solvers.
Experiment Setup Yes We used two choices for the regularizer: λ = 10 1 and λ = 10 3.