adjusted text label positions and AUROC cbar range (#36)

* adjusted text label positions and AUROC cbar range

* unifed HDI to 80%, switched to median AUROC so that it sits closer to the middle of said hdi

* Figure 2 tweaks

* Figure 3 tweaks

* 50% hdpi and median everywhere

* sizing

fig_settings.py

@@ -13,9 +13,11 @@
 # color styling
 def set_colors(n_colors=2):
     global cmap
+    global auroc_cmap
     global pallette
     cmap = "cmr.redshift"
     qualitative_cmap = cmr.get_sub_cmap(cmap, 0.2, 0.8, N=n_colors)
+    auroc_cmap = cmr.get_sub_cmap(cmap, 0.5, 1)
 
     pallette = qualitative_cmap.colors
     mpl.rcParams["axes.prop_cycle"] = mpl.cycler(color=pallette)

plot_fig1.py

@@ -35,8 +35,8 @@
 ax1.set_position([0, 0.4, 0.45, 0.75])
 
 ax1.text(
-    0.025,
-    0.775,
+    0.01,
+    0.79,
     "(a)",
     transform=ax1.transAxes,
     fontsize=13,
@@ -111,7 +111,7 @@
 ax2 = fig.add_subplot(gs[1])
 ax2.text(
     -0.39,
-    1.1,
+    1.13,
     "(b)",
     transform=ax2.transAxes,
     fontsize=12,
@@ -140,7 +140,7 @@
 
 err_c1 = np.zeros((2, n))
 for i in range(n):
-    interval = az.hdi(c1_samples[:, i], hdi_prob=0.95)
+    interval = az.hdi(c1_samples[:, i], hdi_prob=0.5)
     x, y = interval
     err_c1[0, i] = max(c1_mean[i] - x, 0)
     err_c1[1, i] = max(y - c1_mean[i], 0)
@@ -165,7 +165,7 @@
 
 err_c2 = np.zeros((2, n))
 for i in range(n):
-    interval = az.hdi(c2_samples[:, i], alpha=0.05, roundto=4)
+    interval = az.hdi(c2_samples[:, i], hdi_prob=0.5)
     x, y = interval
     err_c2[0, i] = max(c2_mean[i] - x, 0)
     err_c2[1, i] = max(y - c2_mean[i], 0)
@@ -203,8 +203,8 @@
 """
 ax3 = fig.add_subplot(gs[2])
 ax3.text(
-    -0.35,
-    1.05,
+    -0.37,
+    1.11,
     "(c)",
     transform=ax3.transAxes,
     fontsize=12,
@@ -219,12 +219,12 @@
 tmax = np.array(data["tmax"], dtype=float)
 performance = np.array(data[measure], dtype=float)
 
-ax3.semilogx(tmax, performance[0].mean(axis=1), color="C0", label="Simple")
-ax3.semilogx(tmax, performance[1].mean(axis=1), color="C1", label="Complex")
-min_idx = np.where((performance[0].mean(axis=1) - performance[1].mean(axis=1)) < 0)[
+ax3.semilogx(tmax, np.median(performance[0], axis=1), color="C0", label="Simple")
+ax3.semilogx(tmax, np.median(performance[1], axis=1), color="C1", label="Complex")
+min_idx = np.where((np.median(performance[0], axis=1) - np.median(performance[1], axis=1)) < 0)[
     0
 ].min()
-max_idx = np.where((performance[0].mean(axis=1) - performance[1].mean(axis=1)) < 0)[
+max_idx = np.where((np.median(performance[0], axis=1) - np.median(performance[1], axis=1)) < 0)[
     0
 ].max()
 print(tmax[min_idx])
@@ -233,7 +233,7 @@
 hdi_a = np.zeros_like(tmax)
 hdi_b = np.zeros_like(tmax)
 for i in range(len(tmax)):
-    interval = az.hdi(performance[0, i], hdi_prob=0.95)
+    interval = az.hdi(performance[0, i], hdi_prob=0.5)
     a, b = interval
     hdi_a[i] = a
     hdi_b[i] = b
@@ -243,14 +243,14 @@
 hdi_a = np.zeros_like(tmax)
 hdi_b = np.zeros_like(tmax)
 for i in range(len(tmax)):
-    interval = az.hdi(performance[1, i], hdi_prob=0.95)
+    interval = az.hdi(performance[1, i], hdi_prob=0.5)
     a, b = interval
     hdi_a[i] = a
     hdi_b[i] = b
 
 ax3.fill_between(tmax, hdi_a, hdi_b, alpha=0.3, color="C1", edgecolor="none")
 ax3.set_ylabel(measure.upper())
-ax3.set_xlabel(r"$t_{\mathregular{max}}$")
+ax3.set_xlabel(r"Time series length, $T$")
 ax3.set_xlim([10, 10**4])
 ax3.set_xticks(
     [10, 100, 1000, 10000],
@@ -261,8 +261,8 @@
         r"$\mathregular{10^4}$",
     ],
 )
-ax3.set_ylim([0, 1])
-ax3.set_yticks([0, 0.5, 1], [0, 0.5, 1])
+ax3.set_ylim([0.45, 1])
+ax3.set_yticks([0.5, 1], [0.5, 1])
 
 ax3.legend(
     loc="lower right",
@@ -280,7 +280,7 @@
 ax4 = fig.add_subplot(gs[3])
 ax4.text(
     -0.38,
-    1.05,
+    1.11,
     "(d)",
     transform=ax4.transAxes,
     fontsize=12,
@@ -300,20 +300,20 @@
     np.fliplr(to_imshow_orientation(mean_performance)),
     extent=(min(frac), max(frac), max(beta), min(beta)),
     aspect="auto",
-    cmap=cmap,
-    vmin=0,
+    cmap=fs.auroc_cmap,
+    vmin=0.5,
     vmax=1,
 )
-ax4.set_xlabel(r"Complexity, $\lambda$")
+ax4.set_xlabel(r"Complexity, $\omega$")
 ax4.set_ylabel(r"Infectivity, $\beta$")
 
 ax4.set_xticks([0, 0.5, 1], [0, 0.5, 1])
 ax4.set_yticks([0, 0.5, 1], [0, 0.5, 1])
 
 cbar_ax = fig.add_axes([0.91, 0.15, 0.015, 0.31])  # x, y, width, height
 cbar = plt.colorbar(c, cax=cbar_ax)
-cbar.set_label(measure.upper(), fontsize=axislabel_fontsize, rotation=270, labelpad=10)
-cbar_ax.set_yticks([0, 1], [0, 1], fontsize=tick_fontsize)
+cbar.set_label(measure.upper(), fontsize=axislabel_fontsize, rotation=270, labelpad=0)
+cbar_ax.set_yticks([0.5, 1], [0.5, 1], fontsize=tick_fontsize)
 
 plt.savefig("Figures/Fig1/fig1.png", dpi=1000)
 plt.savefig("Figures/Fig1/fig1.pdf", dpi=1000)

plot_fig2.py

@@ -31,7 +31,7 @@
 ]
 keys = ["p", "alpha", "size"]
 titles = ["Erdös-Rényi", "Power-law CM", "Clustered"]
-labels = [r"$p$", r"$\alpha$", r"$s$"]
+labels = [r"Density, $p$", r"Exponent $\alpha$", r"Clique size $s$"]
 xticks = [
     [0, 0.5, 1],
     [-4, -3, -2],
@@ -116,7 +116,7 @@ def visualize_networks(i, ax):
             ax.plot(var, r0 * beta_c, "-", linewidth=0.25, color="grey")
 
     if i == 0:
-        ax.set_ylabel(r"$\beta$")
+        ax.set_ylabel(r"Infectivity, $\beta$")
     else:
         ax.set_yticks([], [])
 
@@ -144,7 +144,7 @@ def visualize_networks(i, ax):
     ax.set_yticks([0, 0.5, 1], [0, 0.5, 1])
 
     if i == 0:
-        ax.set_ylabel(r"$\beta$")
+        ax.set_ylabel(r"Infectivity,$\beta$")
     else:
         ax.set_yticks([], [])
 

plot_fig3.py

@@ -9,16 +9,8 @@
 import fig_settings as fs
 from lcs import *
 
-axislabel_fontsize = 12
-tick_fontsize = 12
-fs.set_fonts(
-    {
-        "font.family": "serif",
-        "axes.labelsize": axislabel_fontsize,
-        "xtick.labelsize": tick_fontsize,
-        "ytick.labelsize": tick_fontsize,
-    }
-)
+
+fs.set_fonts()
 
 with open(f"Data/zkc_tmax_comparison.json") as file:
     data = json.load(file)
@@ -52,7 +44,7 @@
 
 x = tmax
 
-plt.figure(figsize=(5.5, 5))
+plt.figure(figsize=(4, 3.5))
 
 core_values = np.unique(coreness)
 
@@ -61,7 +53,7 @@
     y = node_performance_complex[coreness == k] - node_performance_simple[coreness == k]
     # combine first and last axes.
     y = np.array([y[i, :, j] for i in range(n_k) for j in range(n_r)]).T
-    ymean = y.mean(axis=1)
+    ymean = np.median(y, axis=1)
 
     hdpi_a = np.zeros_like(tmax)
     hdpi_b = np.zeros_like(tmax)
@@ -77,7 +69,7 @@
         markersize=ms,
         color=clist[idx],
         alpha=alpha,
-        label=f"{int(k)}-core nodes",
+        label=f"{int(k)}-core",
     )
     plt.fill_between(x, hdpi_a, hdpi_b, color=clist[idx], alpha=0.1)
 
@@ -98,7 +90,7 @@
 
 plt.legend(frameon=False, loc="upper left")
 plt.ylabel(r"$\Delta\,\phi_i$")
-plt.xlabel(r"$t_{\max}$")
+plt.xlabel(r"$T$")
 sns.despine()
 plt.tight_layout()