Optimal and Adaptive Algorithms for Online Boosting

Authors: Alina Beygelzimer, Satyen Kale, Haipeng Luo

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

Reproducibility Variable Result LLM Response
Research Type Experimental While the focus of this paper is a theoretical investigation of online boosting, we have also performed experiments to evaluate our algorithms. We extended the Vowpal Wabbit open source machine learning system (VW) to include the algorithms studied in this paper. All experiments were done on a diverse collection of 13 publically available datasets. Reported is the 0-1 loss on the test set. The results obtained are given in Figure 4. As can be seen, for most datasets, Online BBM.W had the best performance.
Researcher Affiliation Collaboration Alina Beygelzimer BEYGEL@YAHOO-INC.COM Yahoo Labs, New York, NY 10036 Satyen Kale SATYEN@YAHOO-INC.COM Yahoo Labs, New York, NY 10036 Haipeng Luo HAIPENGL@CS.PRINCETON.EDU Department of Computer Science, Princeton University, Princeton, NJ 08540
Pseudocode Yes Algorithm 1 Online BBM... Algorithm 2 Ada Boost.OL
Open Source Code No The paper mentions extending Vowpal Wabbit (VW) and provides a URL to the general VW GitHub repository (https://github.com/JohnLangford/vowpal_wabbit/). It does not explicitly state that the authors' specific extensions or source code for the described algorithms are publicly released or available at this link.
Open Datasets Yes All experiments were done on a diverse collection of 13 publically available datasets, described in Appendix E in the supplementary material.
Dataset Splits Yes For each dataset, we performed a random split with 80% of the data used for singlepass training and the remaining 20% for testing. We tuned the learning rate, the number of weak learners, and the edge parameter γ (for all but the two versions of Ada Boost.OL) using progressive validation 0-1 loss on the training set. Progressive validation is a standard online validation technique, where each training example is used for testing before it is used for updating the model (Blum et al., 1999).
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., GPU/CPU models, memory, or cloud instances) used for running the experiments.
Software Dependencies No The paper states: 'We extended the Vowpal Wabbit open source machine learning system (VW) to include the algorithms studied in this paper.' It provides a URL for VW but does not specify a version number for Vowpal Wabbit or any other software dependencies.
Experiment Setup Yes We tuned the learning rate, the number of weak learners, and the edge parameter γ (for all but the two versions of Ada Boost.OL) using progressive validation 0-1 loss on the training set.