Simultaneous Model Selection and Optimization through Parameter-free Stochastic Learning

Authors: Francesco Orabona

NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Optimal rates of convergence are proved under standard smoothness assumptions on the target function as well as preliminary empirical results. and Even if this is mainly a theoretical work, we believe that it might also have a big potential in the applied world. Hence, as a proof of concept on the potentiality of this method we have also run a few preliminary experiments, to compare the performance of Pi STOL to an SVM using 5-folds cross-validation to select the regularization weight parameter.
Researcher Affiliation Collaboration Francesco Orabona Yahoo! Labs New York, USA francesco@orabona.com Work done mainly while at Toyota Technological Institute at Chicago.
Pseudocode Yes Algorithm 3 Pi STOL: Parameter-free STOchastic Learning.
Open Source Code No No explicit statement or link to the open-source code for the methodology described in this paper was found. The paper mentions using 'LIBSVM' which is a third-party tool.
Open Datasets Yes Datasets available at http://www.csie.ntu.edu.tw/ cjlin/libsvmtools/datasets/.
Dataset Splits Yes to compare the performance of Pi STOL to an SVM using 5-folds cross-validation to select the regularization weight parameter.
Hardware Specification No No specific hardware details (like GPU/CPU models, memory, or cloud instance types) used for running experiments were mentioned in the paper.
Software Dependencies No The latest version of LIBSVM was used to train the SVM [10].
Experiment Setup Yes to compare the performance of Pi STOL to an SVM using 5-folds cross-validation to select the regularization weight parameter. The experiments were repeated with 5 random shuffles, showing the average and standard deviations over three datasets. ... Pi STOL closely tracks the performance of the tuned SVM when a Gaussian kernel is used. ... Note that in the case of News20, a linear kernel is used over the vectors of size 1355192. ... our unoptimized Matlab implementation of Pi STOL