[WIP] RGE Fisher for Wilson Coefficients

src/smefit/analyze/fisher.py

@@ -62,10 +62,30 @@
 
     """
 
-    def __init__(self, coefficients, datasets, compute_quad):
+    def __init__(self, coefficients, datasets, compute_quad, rgemat=None):
         self.coefficients = coefficients
         self.free_parameters = self.coefficients.free_parameters.index
         self.datasets = datasets
+        self.rgemat = rgemat
+
+        # find name of the operators that are constrained
+        self.constrained_coeffs = [
+            name for name in self.coefficients.name if name not in self.free_parameters
+        ]
+        self.constrained_coeffs_idx = [
+            idx
+            for idx, name in enumerate(self.coefficients.name)
+            if name not in self.free_parameters
+        ]
+
+        # remove the constrained coefficient from the rgemat both in axis 1 and 2
+        if rgemat is not None:
+            self.rgemat = np.delete(
+                self.rgemat, [idx for idx in self.constrained_coeffs_idx], axis=1
+            )
+            self.rgemat = np.delete(
+                self.rgemat, [idx for idx in self.constrained_coeffs_idx], axis=2
+            )
 
         # update eft corrections with the constraints
         if compute_quad:
@@ -77,6 +97,9 @@
             self.new_LinearCorrections = impose_constrain(
                 self.datasets, self.coefficients
             )
+            self.new_LinearCorrectionsNoRGE = impose_constrain(
+                self.datasets, self.coefficients, norge=True
+            ).T
 
         self.lin_fisher = None
         self.quad_fisher = None
@@ -97,6 +120,29 @@
             fisher_tab, index=self.datasets.ExpNames, columns=self.free_parameters
         )
 
+    def compute_wc_fisher(self):
+        fisher_tab = []
+        operators = [
+            name
+            for name in self.datasets.OperatorsNames
+            if name not in self.constrained_coeffs
+        ]
+
+        for i, op in enumerate(operators):
+            idxs = slice(i, i + 1)
+            sliced_rgemat = self.rgemat[:, idxs, :]
+            sigma = np.einsum(
+                "ij, ijk -> ik",
+                self.new_LinearCorrectionsNoRGE[:, idxs],
+                sliced_rgemat,
+            )
+            fisher_row = np.diag(sigma.T @ self.datasets.InvCovMat @ sigma)
+            fisher_tab.append(fisher_row)
+
+        self.wc_fisher = pd.DataFrame(
+            fisher_tab, index=operators, columns=self.free_parameters
+        )
+
     def compute_quadratic(self, posterior_df, smeft_predictions):
         """Compute quadratic Fisher information."""
         quad_fisher = []
@@ -531,12 +577,15 @@
         other=None,
         summary_only=True,
         figsize=(11, 15),
         labels=None,
         column_names=None,
+        wc_fisher=False,
+        wc_to_latex=None,
     ):
-
         fisher_df = self.summary_table if summary_only else self.lin_fisher
         quad_fisher_df = self.summary_HOtable if summary_only else self.quad_fisher
+        if wc_fisher:
+            wc_fisher_df = self.wc_fisher
 
         if other is not None:
 
@@ -570,7 +619,6 @@
                 quad_fisher_dfs = [
                     quad_fisher_df[latex_names.index.get_level_values(level=1)]
                 ]
-
         # reshuffle column name ordering
         if column_names is not None:
             custom_ordering = [list(column.keys())[0] for column in column_names]
@@ -595,36 +643,61 @@
         else:
             ax = plt.gca()
 
-        self.plot_values(ax, fisher_dfs, cmap, norm, labels)
+        if wc_fisher:
+            wc_fisher_dfs = [
+                wc_fisher_df.loc[
+                    latex_names.index.get_level_values(level=1),
+                    latex_names.index.get_level_values(level=1),
+                ]
+            ]
 
-        self.set_ticks(
-            ax,
-            np.arange(fisher_dfs[0].shape[1]),
-            np.arange(fisher_dfs[0].shape[0]),
-            latex_names,
-            x_labels,
-        )
-        ax.set_title(r"\rm Linear", fontsize=20, y=-0.08)
-        cax1 = make_axes_locatable(ax).append_axes("right", size="5%", pad=0.5)
-        colour_bar = fig.colorbar(mpl.cm.ScalarMappable(norm=norm, cmap=cmap), cax=cax1)
+            self.plot_values(ax, wc_fisher_dfs, cmap, norm)
 
-        if quad_fisher_df is not None:
-            ax = fig.add_subplot(122)
-            self.plot_values(ax, quad_fisher_dfs, cmap, norm, labels)
+            self.set_ticks(
+                ax,
+                np.arange(wc_fisher_df.shape[1]),
+                np.arange(wc_fisher_df.shape[0]),
+                latex_names,
+                [wc_to_latex[op] for op in wc_fisher_dfs[0].index],
+            )
+            ax.set_title(r"\rm RGE Fisher Information", fontsize=25)
+            cax1 = make_axes_locatable(ax).append_axes("right", size="5%", pad=0.5)
+            colour_bar = fig.colorbar(
+                mpl.cm.ScalarMappable(norm=norm, cmap=cmap), cax=cax1
+            )
+        else:
+            self.plot_values(ax, fisher_dfs, cmap, norm)
 
             self.set_ticks(
                 ax,
-                np.arange(quad_fisher_dfs[0].shape[1]),
-                np.arange(quad_fisher_dfs[0].shape[0]),
+                np.arange(fisher_dfs[0].shape[1]),
+                np.arange(fisher_dfs[0].shape[0]),
                 latex_names,
                 x_labels,
             )
-            ax.set_title(r"\rm Quadratic", fontsize=20, y=-0.08)
+            ax.set_title(r"\rm Linear", fontsize=20, y=-0.08)
             cax1 = make_axes_locatable(ax).append_axes("right", size="5%", pad=0.5)
             colour_bar = fig.colorbar(
                 mpl.cm.ScalarMappable(norm=norm, cmap=cmap), cax=cax1
             )
 
+            if quad_fisher_df is not None:
+                ax = fig.add_subplot(122)
+                self.plot_values(ax, quad_fisher_dfs, cmap, norm)
+
+                self.set_ticks(
+                    ax,
+                    np.arange(quad_fisher_dfs[0].shape[1]),
+                    np.arange(quad_fisher_dfs[0].shape[0]),
+                    latex_names,
+                    x_labels,
+                )
+                ax.set_title(r"\rm Quadratic", fontsize=20, y=-0.08)
+                cax1 = make_axes_locatable(ax).append_axes("right", size="5%", pad=0.5)
+                colour_bar = fig.colorbar(
+                    mpl.cm.ScalarMappable(norm=norm, cmap=cmap), cax=cax1
+                )
+
         fig.subplots_adjust(top=0.9)
 
         colour_bar.set_label(

src/smefit/analyze/pca.py

@@ -150,7 +150,7 @@ def make_sym_matrix(vals, n_op):
     return m
 
 
-def impose_constrain(dataset, coefficients, update_quad=False):
+def impose_constrain(dataset, coefficients, update_quad=False, norge=False):
     """Propagate coefficient constraint into the theory tables.
 
     Note: only linear contraints are allowed in this method.
@@ -187,7 +187,14 @@ def impose_constrain(dataset, coefficients, update_quad=False):
         temp_coeffs.set_constraints()
 
         # update linear corrections
-        new_linear_corrections.append(temp_coeffs.value @ dataset.LinearCorrections.T)
+        if norge:
+            new_linear_corrections.append(
+                temp_coeffs.value @ dataset.LinearCorrectionsNoRGE.T
+            )
+        else:
+            new_linear_corrections.append(
+                temp_coeffs.value @ dataset.LinearCorrections.T
+            )
 
         # update quadratic corrections, this will include some double counting in the mixed corrections
         if update_quad:

src/smefit/analyze/report.py

@@ -19,7 +19,7 @@
 _logger = logging.getLogger(__name__)
 
 
 class Report:
     r"""Report class manager.
 
     If :math:`\chi^2`, Fisher or Data vs Theory plots are produced it computes the
@@ -79,6 +79,12 @@
         self.data_info = self._load_grouped_data_info(report_config["data_info"])
         # Loads coefficients grouped with latex name
         self.coeff_info = self._load_grouped_coeff_info(report_config["coeff_info"])
+
+        self.coeff_to_latex = {}
+        for _, entries in report_config["coeff_info"].items():
+            for val in entries:
+                self.coeff_to_latex[val[0]] = val[1]
+
         self.html_index = ""
         self.html_content = ""
 
@@ -525,11 +531,17 @@
         fishers = {}
         for fit in fit_list:
             compute_quad = fit.config["use_quad"]
-            fisher_cal = FisherCalculator(fit.coefficients, fit.datasets, compute_quad)
+            fisher_cal = FisherCalculator(
+                fit.coefficients, fit.datasets, compute_quad, fit.rgemat
+            )
             fisher_cal.compute_linear()
+            fisher_cal.compute_wc_fisher()
             fisher_cal.lin_fisher = fisher_cal.normalize(
                 fisher_cal.lin_fisher, norm=norm, log=log
             )
+            fisher_cal.wc_fisher = fisher_cal.normalize(
+                fisher_cal.wc_fisher, norm=norm, log=log
+            )
             fisher_cal.summary_table = fisher_cal.groupby_data(
                 fisher_cal.lin_fisher, self.data_info, norm, log
             )
@@ -562,6 +574,8 @@
                     self.coeff_info,
                     f"{self.report}/fisher_heatmap_{fit.name}",
                     title=title,
+                    wc_fisher=True,
+                    wc_to_latex=self.coeff_to_latex,
                     **fit_plot,
                 )
                 figs_list.append(f"fisher_heatmap_{fit.name}")

src/smefit/loader.py

@@ -22,6 +22,7 @@
         "SMTheory",
         "OperatorsNames",
         "LinearCorrections",
+        "LinearCorrectionsNoRGE",
         "QuadraticCorrections",
         "ExpNames",
         "NdataExp",
@@ -733,6 +734,10 @@ def load_datasets(
         lin_corr_list, n_data_tot, sorted_keys, rgemat
     )
 
+    lin_corr_values_norge = construct_corrections_matrix_linear(
+        lin_corr_list, n_data_tot, sorted_keys
+    )
+
     if use_quad:
         quad_corr_values = construct_corrections_matrix_quadratic(
             quad_corr_list, n_data_tot, sorted_keys, rgemat
@@ -766,6 +771,7 @@ def load_datasets(
         np.array(sm_theory),
         sorted_keys,
         lin_corr_values,
+        lin_corr_values_norge,
         quad_corr_values,
         np.array(exp_name),
         np.array(n_data_exp),