Adaptive SVRG Methods under Error Bound Conditions with Unknown Growth Parameter
Authors: Yi Xu, Qihang Lin, Tianbao Yang
NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we consider some applications in machine learning and present some experimental results. We conduct some experiments to demostrate the effectiveness of the proposed algorithms on several tasks, including ℓ1 regularized squared hinge loss minimization, ℓ1 regularized logistic loss minimization for linear classification problems; and ℓ1 constrained ℓp norm regression, ℓ1 regularized square loss minimization and ℓ1 regularized Huber loss minimization for linear regression problems. We use three datasets from libsvm website: Adult (n = 32561, d = 123), E2006-tfidf (n = 16087, d = 150360), and Year Prediction MSD (n = 51630, d = 90). In each plot, the difference between objective value and optimum is presented in log scale. |
| Researcher Affiliation | Academia | Yi Xu , Qihang Lin , Tianbao Yang Department of Computer Science, The University of Iowa, Iowa City, IA 52242, USA Department of Management Sciences, The University of Iowa, Iowa City, IA 52242, USA {yi-xu, qihang-lin, tianbao-yang}@uiowa.edu |
| Pseudocode | Yes | Algorithm 1 SVRG method under HEB (SVRGHEB(x0, T1, R, θ)) |
| Open Source Code | No | No explicit statement or link for open-source code release for the described methodology is found in the paper. |
| Open Datasets | Yes | We use three datasets from libsvm website: Adult (n = 32561, d = 123), E2006-tfidf (n = 16087, d = 150360), and Year Prediction MSD (n = 51630, d = 90). |
| Dataset Splits | No | The paper mentions using specific datasets and a 'testing set' for one, but it does not explicitly provide training/validation/test split percentages or sample counts for all datasets, nor does it refer to specific predefined split methods with citations. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory amounts, or cloud instance types) used for running the experiments. |
| Software Dependencies | No | The paper does not provide specific software dependencies or library versions (e.g., 'Python 3.8', 'PyTorch 1.9') used for the experiments. |
| Experiment Setup | Yes | For all algorithms, the step size is best tuned. The initial epoch length of SVRG++ is set to n/4 following the suggestion in [2], and the same initial epoch length is also used in our algorithms. We set the regularization parameter of ℓ1 norm and the upper bound of ℓ1 constraint to be 10 4 and 100, respectively. |