Inexact trust-region algorithms on Riemannian manifolds

Authors: Hiroyuki Kasai, Bamdev Mishra

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical evaluations demonstrate that the proposed algorithms outperform state-of-the-art Riemannian deterministic and stochastic gradient algorithms across different applications.
Researcher Affiliation Collaboration Hiroyuki Kasai The University of Electro-Communications Japan kasai@is.uec.ac.jp Bamdev Mishra India bamdevm@microsoft.com
Pseudocode Yes Algorithm 1 Inexact Riemannian trust-region (Inexact RTR) algorithm
Open Source Code Yes The implementation of the proposed algorithms uses the MATLAB toolbox Manopt [40] and is available at https://github.com/hiroyuki-kasai/Subsampled-RTR.
Open Datasets Yes We use three real-world datasets: Yale B [46], COIL-100 [47], and CIFAR-100 [48]... two real-world datasets with r = 10, where Case P3 deals with the MNIST dataset [52]... deals with the Covertype dataset [53]... Jester dataset 1 [54]
Dataset Splits No The paper does not explicitly mention training/validation/test dataset splits, percentages, or absolute counts for validation.
Hardware Specification Yes All simulations are performed in MATLAB on a 4.0 GHz Intel Core i7 machine with 32 GB RAM.
Software Dependencies No The paper mentions 'MATLAB toolbox Manopt' but does not provide specific version numbers for either MATLAB or Manopt.
Experiment Setup Yes By following [38], we set |Sg| = n/10 and |SH| = n/102 except Cases P5, P6, M4, and M5. We set the batch-size to n/10 in RSVRG. All simulations are performed in MATLAB on a 4.0 GHz Intel Core i7 machine with 32 GB RAM.