VENI VINDy VICI Resource Data

Overview This dataset contains preprocessed simulation data for two exemplary reduced-order modeling applications presented in the paper “VENI, VINDy, VICI: A generative reduced-order modeling framework with uncertainty quantification” (Neural Networks). Both datasets include high-dimensional state variables transformed via principal component analysis (PCA), their derivatives, and associated metadata for training and testing machine learning models.The data are suitable for applications in reduced-order modeling, system identification, surrogate modeling, and machine learning-based dynamics prediction. Both files are provided as NumPy binary arrays (`.npy` format) and can be loaded directly into Python-based scientific computing frameworks. Datasets 1. Reaction-Diffusion (reaction_diffusion_data.npy) Description: Simulation results from a two-dimensional reaction-diffusion system that generates spiral wave dynamics. Contents: x: state variables (n_sims, n_timesteps, Nx, Ny, channels) t: Time steps (n_timesteps) 2. MEMS Beam Resonator (mems.npy) Description: Simulation results for a straight double-clamped MEMS (micro-electro-mechanical systems) beam resonator excited at resonance. The model describes the nonlinear dynamical response of a polysilicon beam device. This dataset contains preprocessed and reduced-order representations of beam dynamics data. The data was generated and processed for use in reduced-order modeling, system identification, and machine learning applications. It includes principal component analysis (PCA)-transformed state variables, their derivatives, and reference coordinates. The dataset is suitable for tasks such as dynamic system analysis, surrogate modeling, and control design. Contents: x: PCA-transformed state variables (n_samples, n_pod) dxdt: First derivatives of the PCA-transformed state variables (n_samples, n_pod) dxddt: Second derivatives of the PCA-transformed state variables (n_samples, n_pod) x_test, dxdt_test, dxddt_test: Test set equivalents of the above variables (n_samples_test, n_pod) V: PCA components (n_features, n_pod). ref_coords: Reference coordinates of the beam (n_nodes, 3) params, params_test: Simulation parameters for training and testing (n_samples, n_params), (n_samples_test, n_params) t, t_test: Time steps for training and testing. (n_timesteps), (n_timesteps_test) n_sims, n_timesteps: Number of simulations and timesteps in the dataset. with n_samples=n_sims*n_timesteps, n_features=n_nodes*3. Usage: The dataset is stored as a Python pickle file and can be loaded using the pickle module. It is compatible with Python-based machine learning frameworks such as PyTorch and TensorFlow. License: Please refer to the accompanying license file for terms of use. Citation: If you use this dataset in your research, please cite the associated publication.