When are dynamical systems learned from time series data statistically accurate?

Authors: Jeongjin Park, Nicole Yang, Nisha Chandramoorthy

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

Reproducibility Variable Result LLM Response
Research Type Experimental We verify our results on a number of ergodic chaotic systems and neural network parameterizations, including MLPs, Res Nets, Fourier Neural layers, and RNNs.
Researcher Affiliation Academia Jeongjin (Jayjay) Park School of Computational Science and Engineering Georgia Institute of Technology Atlanta, GA 30332 jpark3141@gatech.edu Nicole Tianjiao Yang Department of Mathematics Emory University Atlanta, GA 30322 tianjiao.yang@emory.edu Nisha Chandramoorthy Department of Statistics The University of Chicago Chicago, IL 60637 nishac@uchicago.edu
Pseudocode No The paper describes algorithms and methods but does not include explicit pseudocode blocks or algorithm listings.
Open Source Code Yes The Python code is available at https://github.com/ni-sha-c/stacNODE.
Open Datasets Yes The first 10,000 data points are used as the training data and the last 8,000 points are the test data.
Dataset Splits No The paper mentions training data and test data, but does not explicitly state a separate validation set or how it was used for hyperparameter tuning if it existed.
Hardware Specification Yes Numerical experiments were conducted using Tesla A100 GPUs with 80GB and 40GB memory capacities.
Software Dependencies No The paper mentions the 'torchdiffeq3 library' and 'Py Torch library' for Adam W optimization, but does not specify version numbers for these software components.
Experiment Setup Yes We use the Runge-Kutta 4-stage time integrator with a time step size of 0.01 to define the map F. ... Our Neural ODE map, Fnn, is learned to approximate F by solving the above optimization with n = 10, 000 training points along an orbit. ... Table 2: Hyperparameter choices (Chaotic Systems, Epochs, Time step, Hidden layer width, Layers, Train, Test size, Neural Network, λ in (3)).