CODES Configuration

Name of Configuration

Provide a unique identifier for this training run (e.g., "my_run123").

Training ID

Dataset

Specify which dataset to use, how to transform it, and any additional dataset-related parameters.

Dataset name (e.g. 'osu2008' or 'branca24')

Apply log10 transform to data

Normalization

Use optimal model hyperparameters

Tolerance (e.g. 1e-30)

Subset Factor (e.g. 1, 2, ...)

Surrogates to Include

Select which surrogate models to train, and specify their training parameters (batch size, epochs). Check the boxes to enable a model, and fill in the corresponding parameters if needed.

Fully Connected Neural Network

Deep Operator Network

Latent Neural ODE

Latent Polynomial Network

Add model

Compare models with each other

Miscellaneous Settings

Configure devices, verbosity, and random seed for reproducibility.

Devices

Verbose output

Random Seed

Models to Train (Benchmark Tasks)

Enable and configure different benchmarking scenarios such as interpolation, extrapolation, sparse training data, uncertainty quantification, and batch size scaling.

Interpolation

Enable Interpolation

Use every n-th timestep:

Extrapolation

Enable Extrapolation

Cutoff at timestep n:

Sparse Training Data

Enable Sparse Training Data

Take every n-th sample:

Uncertainty Quantification

Enable Uncertainty Quantification

Ensemble size:

Batch Size Scaling

Enable Batch Size Scaling

Batch sizes to evaluate:

Evaluations During Benchmark

Select which evaluations to perform and record during the benchmark.

Save and plot losses

Save and plot dynamic accuracy

Save and plot timing

Measure Compute performance

Configuration

Config Generator