Efficient online learning with kernels for adversarial large scale problems
Authors: Rémi Jézéquel, Pierre Gaillard, Alessandro Rudi
NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we perform in Section 4 several experiments based on real and simulated data to compare the performance (in regret and in time) of our methods with competitors. 4 Experiments We empirically test PKAWV against several state-of-the-art algorithms for online kernel regression. In particular, we test our algorithms in (1) an adversarial setting [see Appendix G], (2) on large scale datasets. The algorithms are evaluated on four datasets from UCI machine learning repository. In particular, casp (regression) and ijcnn1, cod-rna, SUSY (classification) [see Appendix G for casp and ijcnn1] ranging from 4 104 to 6 106 datapoints. For all datasets, we scaled x in [ 1, 1]d and y in [ 1, 1]. In Figs. 2 and 4 we show the average loss (square loss for regression and classification error for classification) and the computational costs of the considered algorithm. |
| Researcher Affiliation | Academia | INRIA Département d Informatique de l École Normale Supérieure PSL Research University, Paris, France |
| Pseudocode | No | The paper describes the PKAWV algorithm and its implementation details (e.g., 'Appendix H details how (4) can be efficiently implemented in these cases.'), but does not contain a clearly labeled pseudocode or algorithm block within the main text or appendices. |
| Open Source Code | Yes | The code necessary to reproduce the following experiments is available on Git Hub at https://github.com/Remjez/kernel-online-learning. |
| Open Datasets | Yes | The algorithms are evaluated on four datasets from UCI machine learning repository. In particular, casp (regression) and ijcnn1, cod-rna, SUSY (classification) [see Appendix G for casp and ijcnn1] ranging from 4 104 to 6 106 datapoints. |
| Dataset Splits | No | The paper describes an online learning setting where data is processed sequentially. It does not provide specific train/validation/test dataset split percentages, sample counts, or explicit splitting methodology for reproducing the data partitioning of the datasets used in experiments (casp, ijcnn1, cod-rna, SUSY). |
| Hardware Specification | Yes | All experiments have been done on a single desktop computer (Intel Core i7-6700) with a timeout of 5-min per algorithm. |
| Software Dependencies | No | The paper states 'The algorithms above have been implemented in python with numpy' but does not provide specific version numbers for Python, numpy, or any other libraries or solvers used. |
| Experiment Setup | Yes | For all algorithms and all experiments, we set σ = 1 [except for SUSY where σ = 4, to match accuracy results from RCR17] and λ = 1. When using KORS, we set µ = 1, β = 1 and ε = 0.5 as in [CLV17b]. The number of random-features in FOGD is fixed to 1000 and the learning rate η is 1/ n. |