priyamdas2/pared: pared v0.1.0

pared v0.1.0

This release provides a citable archived version of pared, an R package for Pareto-based model selection using multi-objective optimization.

pared is designed for model-selection settings in which users wish to balance multiple criteria, such as goodness-of-fit, sparsity, shrinkage, smoothness, structural similarity, and computational cost. The package provides a generic interface for user-defined multi-objective model-selection problems, together with built-in convenience wrappers for elastic net, fused lasso, fused graphical lasso, and group graphical lasso. The optimization routine uses Gaussian-process-based multi-objective optimization to approximate the Pareto front, and the package provides interactive two- and three-dimensional visualizations to help users inspect trade-offs among candidate models.

Main functionality

• Generic Pareto-based model-selection interface through pared_optimize().

• Support for user-defined objective functions, objective names, and objective directions.

• Built-in wrappers for elastic net, fused lasso, group graphical lasso, and fused graphical lasso.

• Interactive visualization of Pareto-optimal model-selection trade-offs.

• Examples demonstrating direct use with custom objectives and objective-space preferences.

• Simulation and case-study examples illustrating variable-selection and network-selection workflows.

Software availability

The development repository is available at:

https://github.com/priyamdas2/pared

This release corresponds to the software version archived for reproducible use with the revised Bioinformatics Advances manuscript.