qpots.tsemo_runner

Python wrapper around the bundled MATLAB TS-EMO implementation.

class qpots.tsemo_runner.TSEMORunner(func: str, x: list, y: list, lb: list, ub: list, iters: int, batch_number: int)[source]

Bases: object

Runs the TS-EMO algorithm iteratively using MATLAB.

The TSEMORunner class interfaces with MATLAB’s TS-EMO implementation, allowing users to iteratively optimize a multi-objective function while updating results at each step.

Notes

  • Requires a working MATLAB installation and the MATLAB Engine API for Python.

  • Paths to the TS-EMO MATLAB files must be correctly configured.

tsemo_hypervolume(Y: Tensor, ref_point: Tensor, train_shape: int, iters: int)[source]

Compute the hypervolume and Pareto front for a given set of objective values.

This function applies the Fast Nondominated Partitioning algorithm to evaluate hypervolume improvement over multiple iterations.

Parameters:

  • Y (torch.Tensor) – A tensor of objective values, where each row represents a solution’s evaluated objectives.

  • ref_point (torch.Tensor) – A reference point for hypervolume calculation, typically set to be worse than the worst observed objective values.

  • train_shape (int) – The number of initial training points. Determines how many points are included in the hypervolume calculation at each step.

  • iters (int) – The number of iterations the optimization was run for.

Returns:

  • hv: A list containing the hypervolume values computed at each iteration.

  • pf: A tensor representing the Pareto front (set of nondominated solutions).

Return type:

Tuple[list, torch.Tensor]

tsemo_run(save_dir: str, rep: int)[source]

Run the TS-EMO algorithm iteratively and save results after each iteration.

Parameters:

  • save_dir (str) – The directory where results should be saved.

  • rep (int) – The repetition number used to differentiate saved files.

Returns:

  • X: The updated input design points.

  • Y: The updated objective function evaluations.

  • times: The runtime per iteration.

Return type:

Tuple[np.ndarray, np.ndarray, np.ndarray]

Raises:

matlab.engine.EngineError – If the MATLAB engine encounters an error while running TS-EMO.

exception qpots.tsemo_runner._MissingEngineError[source]

Bases: Exception

qpots.tsemo_runner._missing_start_matlab(*args, **kwargs)[source]