save current state

activ/cca/viz.py

@@ -116,6 +116,6 @@ def cross_decomp_scatter(x, y, regressor=LinearRegression(), labels=None, fitlin
         yfit = regressor.predict(xfit)
 
         ax.plot(xfit, yfit, color='black')
-        x_pos, y_pos = (0.5, 0.1)
+        x_pos, y_pos = (0.7, 0.1)
         ax.text(x_pos, y_pos, "$R^2$ = %0.2f" % cv_r2, size=fontsize, transform=ax.transAxes)
     return ax

activ/clustering/viz.py

@@ -138,7 +138,7 @@ def shade_around(data, ax=None, stretch=None, alpha=0.2, color='blue', convex=Tr
         # zi = interpolator(Xi, Yi)
         # cfset = ax.contourf(xi, yi, zi, colors=color, alpha=alpha)
 
-        cfset = ax.tricontourf(x, y, z, colors=color, alpha=alpha)
+        cfset = ax.tricontourf(x, y, z, colors=color, alpha=alpha, zorder=0)
     else:
         ch = concave_hull(data, 16)
         gpd.GeoSeries([ch]).plot(ax=ax, alpha=alpha, color=color, edgecolor='none')
@@ -459,7 +459,7 @@ def plot_max1d_simdata_results(path, ax=None, flip=False, fontsize='x-large'):
     sim_sweep_plot(est_noc_max1d, true_noc, ax=ax, flip=flip, fontsize=fontsize)
 
 
-def make_clustered_plot(emb, n_clusters, feature_colors, weights=None, ax=None, stretch=None, fs='x-large', **kwargs):
+def make_clustered_plot(emb, n_clusters, feature_colors, weights=None, ax=None, stretch=None, fs='x-large', add_labels=False, highlight=0, **kwargs):
     """
     Plot 2-D UMAP embedding. Cluster embedding and shade around the
     resulting clusters. Pass in weights to plot points as pie charts
@@ -474,6 +474,28 @@ def make_clustered_plot(emb, n_clusters, feature_colors, weights=None, ax=None,
     else:
         cluster_plot(emb, labels, colors=feature_colors, ax=ax, stretch=stretch, fs=fs, **kwargs)
 
+    if add_labels:
+        for label in np.unique(labels):
+            mask = labels == label
+            center = np.mean(emb[mask], axis=0)
+            ax.text(center[0], center[1], str(label), fontsize=fs)
+    elif highlight > 0:
+        if weights is None:
+            raise ValueError("to highlight clusters, please provide weights for determining homogeneity")
+        print(np.unique(labels, return_counts=True))
+        label_var = np.zeros(max(labels) + 1)
+        for label in np.unique(labels):
+            mask = labels == label
+            label_var[label] = np.var(weights[mask], axis=0).sum()
+        print('label_var before', label_var)
+        label_var = np.argsort(label_var)[:highlight]
+        print('label_var after', label_var)
+        for label in label_var:
+            mask = labels == label
+            print(mask.sum())
+            shade_around(emb[mask], ax=ax, color='lightgrey', convex=True)
+
+    return labels
 
 def get_real_noc(tested_noc, foc, smooth=True, use_median=False, spread_asm=True, spread_foc=True):
     ret = dict()
@@ -587,7 +609,7 @@ def plot_real_foc_results(path, ax=None, max1d_cutoff=False, ci=None, n_sigma=1,
 
     ax.set_ylabel(ylabel, fontsize=fontsize)
     ax.set_xlabel("# Outcome clusters", fontsize=fontsize)
-    return est
+    return est, med
 
 
 def plot_real_accuracy_chance_results(path, ax=None, fontsize='x-large', ylabel="Prediction Accuracy"):
@@ -600,16 +622,16 @@ def plot_real_accuracy_chance_results(path, ax=None, fontsize='x-large', ylabel=
 
 
     x_noc, lower, med, upper = flatten_summarize(np.arange(2, 51),
-                                                 real_chance, smooth=False,
+                                                 real_accuracy, smooth=False,
                                                  iqr=True)
-
-    plot_line(x_noc, med, lower=lower, upper=upper, ax=ax, color='gray')
+    plot_line(x_noc, med, lower=lower, upper=upper, ax=ax, color='black', label='Accuracy')
+    raw = med
 
     x_noc, lower, med, upper = flatten_summarize(np.arange(2, 51),
-                                                 real_accuracy, smooth=False,
+                                                 real_chance, smooth=False,
                                                  iqr=True)
-
-    plot_line(x_noc, med, lower=lower, upper=upper, ax=ax, color='black')
+    plot_line(x_noc, med, lower=lower, upper=upper, ax=ax, color='gray', label='Chance accuracy')
+    chance = med
 
     yticks = np.array([2, 4, 6, 8, 10], dtype=np.int)
     ax.set_yticks(yticks/10)
@@ -620,6 +642,7 @@ def plot_real_accuracy_chance_results(path, ax=None, fontsize='x-large', ylabel=
 
     ax.set_ylabel(ylabel, fontsize=fontsize)
     ax.set_xlabel("# Outcome clusters", fontsize=fontsize)
+    return raw, chance
 
 
 def entropy_across_clusterings(variable, labels):

activ/nmf/viz.py

@@ -430,8 +430,8 @@ def plot_umap_nmf_piechart(weights, umap_emb, s=100, ax=None, fontsize=None, pal
     """
     weights_pie_scatter(weights, umap_emb, s=s, ax=ax, palette=palette)
     ax.tick_params(labelsize=fontsize)
-    ax.set_xlabel('UMAP dimesion 1', fontsize=fontsize)
-    ax.set_ylabel('UMAP dimesion 2', fontsize=fontsize)
+    ax.set_xlabel('UMAP dimension 1', fontsize=fontsize)
+    ax.set_ylabel('UMAP dimension 2', fontsize=fontsize)
 
 
 def plot_umap_nmf_max(emb, weights, bases_labels, right=False, min_dist=0.0, legend=True, ax=None, palette=None):

activ/viz.py

@@ -15,6 +15,10 @@
 from .analytics import cv_r2_score, linefit, _check_X_y
 
 
+def remove_ticks(ax):
+    ax.tick_params(axis='x', which='both', bottom=False, top=False, labelbottom=False)
+    ax.tick_params(axis='y', which='both', left=False, right=False, labelleft=False)
+
 
 def get_nmf_colors():
     """