reproducibilityindex.ai

Non-parametric Representation Learning with Kernels

Authors: Pascal Esser, Maximilian Fleissner, Debarghya Ghoshdastidar

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental empirically evaluate the performance of these methods in both small data regimes as well as in comparison with neural network based models.In this section we illustrate the empirical performance of the kernel-based representation learning models introduced in this paper.
Researcher Affiliation Academia Technical University of Munich, Germany esser@cit.tum.de, fleissner@cit.tum.de, ghoshdas@cit.tum.de
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes We provide a Python implementation on https://github.com/ pascalesser/Representation-Learning-with-Kernels.
Open Datasets Yes concentric circles, cubes (Pedregosa et al. 2011), Iris (Fisher 1936) and Ionosphere (Sigillito. et al. 1989), CIFAR-10 (Krizhevsky, Hinton et al. 2009), MNIST (Deng 2012), SVHN (Netzer et al. 2011)
Dataset Splits Yes We fix the following data split: π‘’π‘›π‘™π‘Žπ‘π‘’π‘™π‘™π‘’π‘‘= 50%, π‘™π‘Žπ‘π‘’π‘™π‘™π‘’π‘‘= 5% and 𝑑𝑒𝑠𝑑= 45%, We perform leave-one-out validation on π‘Ώπ‘™π‘Žπ‘. to pick the bandwidth of the method applied to the test set.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions a "Python implementation" but does not specify the version of Python or any other software dependencies with their respective version numbers.
Experiment Setup Yes We fix the following data split: π‘’π‘›π‘™π‘Žπ‘π‘’π‘™π‘™π‘’π‘‘= 50%, π‘™π‘Žπ‘π‘’π‘™π‘™π‘’π‘‘= 5% and 𝑑𝑒𝑠𝑑= 45%, and consider β„Ž= 2 as the embedding dimension., We consider a π‘˜-nn model (with π‘˜= 3), For Gaussian and Laplacian kernel we choose the bandwidth using a grid search over 15 steps spaced logarithmically between 0.01 and 100., noisy version are generated by 𝒙 = 𝒙+ πœ€, πœ€π‘– (0, 0.1).