Tracking Approximate Solutions of Parameterized Optimization Problems over Multi-Dimensional (Hyper-)Parameter Domains

Authors: Katharina Blechschmidt, Joachim Giesen, Soeren Laue

ICML 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results for kernelized support vector machines and the elastic net confirm the theoretical complexity analysis.
Researcher Affiliation Academia Friedrich-Schiller-Universit at Jena, Germany
Pseudocode No Section 4 describes the algorithm's steps in prose ('Initialization.' and 'Iteration.') rather than in a structured pseudocode or algorithm block.
Open Source Code No The paper states 'In our implementation of the adaptive algorithm from Section 4 we used the LIBSVM package, see (Fan et al., 2005), to compute a near optimal dual solution at a given grid vertex...' and 'In our implementation of the adaptive algorithm from Section 4 we used GLMNET, see (Friedman et al., 2010), for solving the primal optimization problem...' but does not provide a link to the authors' own source code for the described methodology.
Open Datasets Yes The data sets that have been used in our experiments were obtained from the LIBSVM website, see (Lin) for a description. (Lin) LIBSVM Tools. Data sets available at www.csie.ntu.edu.tw/~cjlin/ libsvmtools/datasets/.
Dataset Splits Yes In Figure 2(middle) a 10-fold cross-validation plot is shown for the same data set. In Figure 3(middle/left) a 10-fold cross-validation RMSE plot is shown for the same data set.
Hardware Specification No The paper does not provide any specific hardware details (e.g., GPU/CPU models, memory amounts) used for running its experiments.
Software Dependencies No The paper mentions using 'LIBSVM package' and 'GLMNET' but does not specify version numbers for these or any other software dependencies.
Experiment Setup Yes We considered the two-dimensional parameter space (c, γ) with c [2^10, 2^10] and γ [2^10, 2^10], and a uniform grid with vertices at (2^i, 2^j), where i and j were incremented in steps of 0.05, i.e., the grid had 400 x 400 = 160,000 vertices. We considered parameter values λ [0, 1] and c [2^10, 2^5].