Scalars are universal: Equivariant machine learning, structured like classical physics

Authors: Soledad Villar, David W Hogg, Kate Storey-Fisher, Weichi Yao, Ben Blum-Smith

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We complement our theory with numerical examples that show that the scalar-based method is simple, efficient, and scalable. ... We present numerical experiments using our scalar-based approach compared to other methods in Section 7 (see also [96]).
Researcher Affiliation Academia Soledad Villar Department of Applied Mathematics and Statistics Johns Hopkins University David W. Hogg Flatiron Institute a divison of the Simons Foundation Kate Storey-Fisher Center for Cosmology and Particle Physics Department of Physics, New York University Weichi Yao Department of Technology, Operations, and Statistics Stern School of Business, New York University Ben Blum-Smith Center for Data Science New York University
Pseudocode No The paper does not include any pseudocode or algorithm blocks.
Open Source Code Yes The code is available on Git Hub2, and it reuses much of the functionality provided by EMLP [28].
Open Datasets Yes We demonstrate our approach using scalar-based multi-layer perceptrons (MLP) on two toy learning tasks from [28]: an O(5)-invariant task and an O(3)-equivariant task.
Dataset Splits No The paper mentions 'Test error as a function of training set size' but does not specify the splits for training, validation, or test sets in percentages or absolute counts, nor does it explicitly mention a validation set.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments.
Software Dependencies No The paper mentions that the code reuses functionality from EMLP [28] and uses MLPs, but it does not specify software names with version numbers (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup No The paper describes the tasks and models used (MLPs), but it does not provide specific experimental setup details such as learning rates, batch sizes, number of epochs, or optimizer settings.