Temporal Variability in Implicit Online Learning

Authors: Nicolò Campolongo, Francesco Orabona

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we validate our theoretical findings on classification and regression datasets.
Researcher Affiliation Academia Nicol o Campolongo Universit a di Milano nicolo.campolongo@unimi.it Francesco Orabona Boston University francesco@orabona.com
Pseudocode Yes Algorithm 1 Implicit Online Mirror Descent (IOMD) and Algorithm 2 Ada Implicit
Open Source Code No The paper does not provide explicit statements or links for the open-source code of the described methodology.
Open Datasets Yes We used datasets from the LIBSVM library [6].
Dataset Splits No The paper mentions using datasets from the LIBSVM library and that 'Details about the datasets can be found in Appendix D.', but the main text and Appendix D do not specify training/validation/test dataset splits.
Hardware Specification No The paper does not provide any specific details about the hardware used to run the experiments.
Software Dependencies No The paper does not specify software dependencies with version numbers.
Experiment Setup Yes We set β = 1 in all algorithms. ... We consider values of β in [2 20, 220] with a grid containing 41 points. ... For classification tasks we use the hinge loss, while for regression tasks we use the absolute loss. In both cases, we adopt the squared L2 function for ψ. ... Before running the algorithms, we preprocess the data by dividing each feature by its maximum absolute value so that all the values are in the range [ 1, 1], then we add a bias term.