Spherical Random Features for Polynomial Kernels
Authors: Jeffrey Pennington, Felix Xinnan X. Yu, Sanjiv Kumar
NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We compare the SRF method with Random Maclaurin (RM) [14] and Tensor Sketch (TS) [15], the other polynomial kernel approximation approaches. Throughout the experiments, we choose the number of Gaussians, N, to equal 10, though the specific number had negligible effect on the results. The bias term is set as a = 4. Other choices such as a = 2, 3 yield similar performance; results with a variety of parameter settings can be found in the Supplementary Material. The error bars and standard deviations are obtained by conducting experiments 10 times across the entire dataset. |
| Researcher Affiliation | Industry | Jeffrey Pennington Felix X. Yu Sanjiv Kumar Google Research {jpennin, felixyu, sanjivk}@google.com |
| Pseudocode | Yes | Algorithm 1 Spherical Random Fourier (SRF) Features |
| Open Source Code | No | The paper does not provide an explicit statement or a link to the source code for the methodology described. |
| Open Datasets | Yes | We investigate the scalability of the SRF method on the Image Net 2012 dataset, which consists of 1.3 million 256 256 color images from 1000 classes. [...] We train linear classifiers with liblinear [3] and evaluate classification accuracy on various datasets, two of which are summarized in Table 1; additional results are available in the Supplementary Material. (Table 1 mentions 'usps', 'gisette', 'adult' datasets) |
| Dataset Splits | No | The paper mentions using specific datasets for training and evaluation but does not specify exact training, validation, or test split percentages or sample counts in the main text. |
| Hardware Specification | No | The paper mentions 'fixed hardware' in a figure caption but does not provide any specific details about the CPU, GPU, or other hardware components used for the experiments. |
| Software Dependencies | No | The paper mentions 'liblinear [3]' but does not provide a specific version number for this or any other software dependency. |
| Experiment Setup | Yes | The bias term is set as a = 4. Other choices such as a = 2, 3 yield similar performance; results with a variety of parameter settings can be found in the Supplementary Material. [...] We use the same architecture and parameter settings as [22] (including the fixed convolutional neural network parameters), except we replace the RFF kernel layer with an ℓ2 normalization step and an SRF kernel layer with parameters a = 4 and p = 10. |