SFI.utils.formatting module

Pretty-printing utilities for inferred models.

SFI.utils.formatting.model_summary(labels, coefficients, *, stderr=None, support=None, coeffs_true=None, support_true=None, title='Coefficient Table', max_rows=60, significance_thresholds=(2.0, 10.0, 100.0), auto_labels=False)[source]

Build a human-readable coefficient table with SNR and significance.

Only active (support) coefficients are shown in the table body. Zeroed basis functions are listed separately below, unless labels are auto-generated.

Parameters:

  • labels (sequence of str) – One label per basis function.

  • coefficients (1-D array) – Coefficient vector (length must match labels or support).

  • stderr (1-D array or None) – Standard errors (same length as coefficients). These reflect sampling error only; discretization (finite-time-step) bias is not included.

  • support (1-D int array or None) – Indices into labels that coefficients correspond to. If None, full support is assumed (all basis functions have non-zero coefficients).

  • title (str) – Section header printed above the table.

  • max_rows (int) – If the table exceeds this many rows, truncate the middle.

  • significance_thresholds (tuple of float) – Three SNR thresholds (in multiples of stderr) for the *, **, and *** significance levels. Defaults (2.0, 10.0, 100.0).

  • auto_labels (bool) – If True, labels were auto-generated (e.g. b0, b1, …) and the list of zeroed functions is suppressed.

  • coeffs_true (ndarray | None)

  • support_true (ndarray | None)

Returns:

Ready-to-print multi-line table. Terms are marked , *, or * according to *significance_thresholds*; * terms are bold, * and ** terms are normal weight, non-significant active terms are dimmed.

Return type:

str

SFI.utils.formatting.print_model_comparison(inferences, labels, *, extra_cols=None, metrics=None, title='Model Comparison')[source]

Build a multi-model comparison table from several inference objects.

Parameters:

  • inferences – Sequence of fitted inference objects.

  • labels – One name per inference (row labels).

  • metrics – Attribute names to read from each inference (default ["n_params", "NMSE_force", "force_predicted_MSE"]). The special "n_params" counts force_coefficients_full.

  • extra_cols – Optional {column_name: {label: value}} of caller-supplied cells.

  • title (str) – Header line.

Returns:

Ready-to-print table.

Return type:

str