SAGA with Arbitrary Sampling

Authors: Xun Qian, Zheng Qu, Peter Richtárik

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

Reproducibility Variable Result LLM Response
Research Type Experimental We tested SAGA-AS to solve the logistic regression problem (23) on 3 different datasets: w8a, a9a and ijcnn14. The experiments presented in Section 5.1 and 5.2 are tested for λ1 = 0 and λ2 = 1e 5, which is of the same order as the number of samples in the three datasets. In Section 5.3 we test on the unregularized problem with λ1 = λ2 = 0. In all the plots, the x-axis records the number of pass of the dataset. More experiments can be found in the Suppl.
Researcher Affiliation Academia 1King Abdullah University of Science and Technology, Thuwal, Kingdom of Saudi Arabia 2University of Hong Kong, Hong Kong 3Moscow Institute of Physics and Technology, Dolgoprudny, Moscow Region, Russia.
Pseudocode Yes Algorithm 1 SAGA with Arbitrary Sampling (SAGA-AS)
Open Source Code No The paper does not provide any explicit statement about releasing source code for the SAGA-AS method, nor does it include a link to a code repository.
Open Datasets Yes We tested SAGA-AS to solve the logistic regression problem (23) on 3 different datasets: w8a, a9a and ijcnn14. ... 4https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/
Dataset Splits No The paper mentions using specific datasets but does not explicitly describe the training, validation, or test dataset splits (e.g., percentages or sample counts).
Hardware Specification No The paper does not specify any particular hardware used for running the experiments, such as CPU or GPU models, or cloud computing instances.
Software Dependencies No The paper mentions “libsvmtools/datasets/” as a source for datasets but does not provide specific version numbers for any software dependencies, libraries, or programming languages used for the implementation or experiments.
Experiment Setup Yes The experiments presented in Section 5.1 and 5.2 are tested for λ1 = 0 and λ2 = 1e 5, which is of the same order as the number of samples in the three datasets. In Section 5.3 we test on the unregularized problem with λ1 = λ2 = 0. ... We compare uniform sampling SAGA (SAGA-UNI) with importance sampling SAGA (SAGA-IP), as described in Section 3.3 , on three values of τ {1, 10, 50}.