A Mathematical Model for Curriculum Learning for Parities

Authors: Elisabetta Cornacchia, Elchanan Mossel

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In all our experiments we use fully connected ReLU networks and we train them by SGD on the square loss.
Researcher Affiliation Academia 1Institute of Mathematics, EPFL, Lausanne, Switzerland 2Department of Mathematics and IDSS, MIT, US.
Pseudocode No The paper describes mathematical update rules and algorithmic steps within the text, but does not include a formal 'Algorithm' or 'Pseudocode' block.
Open Source Code Yes Code: https://github.com/ecornacchia/Curriculum-Learning-for-Parities
Open Datasets No The paper describes generating samples from specific distributions (e.g., Rad(p)d) rather than using a pre-existing, publicly available dataset with concrete access information or citations.
Dataset Splits No The paper describes training strategies and implicit testing but does not provide explicit training, validation, or test dataset split percentages or sample counts.
Hardware Specification No The paper does not provide specific details about the hardware used for experiments, such as GPU or CPU models.
Software Dependencies No The paper mentions using ReLU networks and training with SGD on square loss, but does not specify versions for software libraries or dependencies within the text.
Experiment Setup Yes In all plots, we use a 2-layers Re LU MLP with batch size 1024, input dimension 100, and 100 hidden units. We run a 2-steps curriculum strategy for 3 values of p1, namely p1 = 39/40, 19/20, 1/20.