Derivatives of Stochastic Gradient Descent in parametric optimization

Authors: Franck Iutzeler, Edouard Pauwels, Samuel Vaiter

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

Reproducibility Variable Result LLM Response
Research Type Experimental Our theoretical findings are illustrated by numerical experiments on synthetic tasks.
Researcher Affiliation Academia Franck Iutzeler Université Paul Sabatier, Institut de Mathématiques de Toulouse, France. Edouard Pauwels Toulouse School of Economics, Université Toulouse Capitole, Toulouse, France. Samuel Vaiter CNRS & Université Côte D Azur, Laboratoire J. A. Dieudonné. Nice, France.
Pseudocode No The paper describes algorithms like SGD and its derivative recursion but does not present them in a formalized pseudocode block or algorithm figure.
Open Source Code Yes The code is released along with the paper.
Open Datasets Yes The dataset used is ijcnn1 from libsvm with 49,990 observations and 22 features.
Dataset Splits No The paper does not explicitly provide details about training, validation, or test dataset splits.
Hardware Specification Yes All the experiments were performed in jax (Bradbury et al., 2018) on a Mac Book Pro M3 Max.
Software Dependencies No The paper mentions that experiments were performed in 'jax (Bradbury et al., 2018)' but does not specify a version number for jax or any other software dependencies like numpy or matplotlib mentioned in the supplementary material.
Experiment Setup Yes We consider various step size regimes and set η0 = µ 4L2 for all experiments. All experiences are performed with m, d = 100, 10 and µ = 0.05.