Changelog

You can now incorporate insights from literature or lab data directly into your models using Bayesian priors—bringing more accuracy and confidence to your parameter estimation and uncertainty quantification.

• Smarter Modeling: Integrate expert knowledge with ease using Bayesian priors for more informed parameter estimation.

• Flexible Prior Types: Choose from normal or lognormal distributions, available in both univariate and multivariate forms—ideal for handling parameter collinearity.

• Better Inference: Achieve more robust and trustworthy results with enhanced support for uncertainty quantification.

• Plug-and-Play: Fully backward compatible and fits seamlessly into existing workflows.

After parameterizing a model, one of the most important steps is validating how well it generalizes to real-world data. Ionworks now makes this step effortless with built-in support for parameter validation in the parameterization pipeline.

Key Features:

• Easy Validation Setup: Pick your model and parameter set, upload one or more validation datasets, and let Ionworks handle the rest.

• Automated Model Runs: The selected model is automatically run against the validation data.

• Fit Quality Report: Quickly see how well the model fits each dataset with clear, visual feedback—no manual scripting required.

This helps you catch overfitting, compare parameter sets, and build confidence in your model before moving on to design or optimization.

Running and managing parameterizations just got a major upgrade. Ionworks now supports a standard configuration format and a command-line interface (CLI) to make large-scale, reproducible parameterization easier than ever.

Key Features:

• Batch Parameterization: Run large sets of parameterizations across multiple datasets with a single config file.

• Experiment Management: Swap models, cost functions, optimizers, and hyperparameters without rewriting code—perfect for rapid iteration.

• Cloud Execution: Launch parameterizations on the cloud seamlessly, with full support for parallel runs.

Configurations can be defined flexibly in JSON, YAML, or Python dictionaries, and executed either:

• Directly from Python, or

• From the Command Line Interface (CLI)

This new tooling makes it easy to scale up your workflows and bring consistency to even the most complex parameterization experiments.