PAC-Bayes bounds for stable algorithms with instance-dependent priors
Authors: Omar Rivasplata, Emilio Parrado-Hernandez, John S. Shawe-Taylor, Shiliang Sun, Csaba Szepesvari
NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The purpose of the experiments was to explore the strengths and potential weaknesses of our new bound in relation to the previous alternatives, as well as to explore the bound s ability to help model selection. For this, to facilitate comparisons, taking the setup of Parrado-Hern andez et al. [2012], we experimented with the five UCI datasets described there. |
| Researcher Affiliation | Collaboration | Emilio Parrado-Hern andez University Carlos III of Madrid, Shiliang Sun East China Normal University, Omar Rivasplata is sponsored by Deep Mind via an Overseas Impact Studentship to undertake grad studies at UCL Department of Computer Science. |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any statement or link for open-source code for its methodology. |
| Open Datasets | Yes | we experimented with the five UCI datasets described there. |
| Dataset Splits | Yes | The datasets were split into a training and a test set using the train test split method of scikit, keeping 80% of the data for training and 20% for testing. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU/GPU models, memory) used for running its experiments. |
| Software Dependencies | No | The paper mentions "a Python implementation of the SMO algorithm" and "scikit" but does not provide specific version numbers for software dependencies. |
| Experiment Setup | Yes | We used an offset-free SVM classifier with a Gaussian RBF kernel (x, y) = exp( kx yk2 / (2σ2)) with RBF width parameter σ > 0. The SVM used the so-called standard SVM-C formulation which multiplies the total (hinge) loss by C > 0;... we set up a geometric 7 x 7 grid over the (C, σ)-parameter space where C ranges between 2^-8C0 and 2^2C0 and σ ranges between 2^-3σ0 and 2^3σ0, where σ0 is the median of the Euclidean distance between pairs of data points of the training set, and given σ0, C0 is obtained as the reciprocal value of the empirical variance of data in feature space underlying the RBF kernel with width σ0. |