prepare for publication

awarememristor/plotting/figures.py

@@ -189,7 +189,7 @@ def SiO_x():
     utils.save_fig(fig, "SiO_x")
 
 
-def HfO2():
+def Ta_HfO2():
     fig, axes = utils.fig_init(2, 0.45, fig_shape=(1, 2), sharey=True)
 
     data = simulations.data.load_Ta_HfO2()
@@ -292,111 +292,7 @@ def HfO2():
         handles=handles,
     )
 
-    utils.save_fig(fig, "HfO2")
-
-
-def pf_residuals(is_d_times_perm: bool = False):
-    fig, axes = utils.fig_init(2, 1.0, fig_shape=(4, 2))
-
-    axes[0, 1].sharey(axes[0, 0])
-    plt.setp(axes[0, 1].get_yticklabels(), visible=False)
-    axes[1, 1].sharey(axes[1, 0])
-    plt.setp(axes[1, 1].get_yticklabels(), visible=False)
-
-    axes[1, 0].sharex(axes[0, 0])
-    plt.setp(axes[0, 0].get_xticklabels(), visible=False)
-    axes[1, 1].sharex(axes[0, 1])
-    plt.setp(axes[0, 1].get_xticklabels(), visible=False)
-
-    axes[3, 0].sharex(axes[2, 0])
-    plt.setp(axes[2, 0].get_xticklabels(), visible=False)
-    axes[3, 0].set_xlim([-2, 2])
-    axes[3, 1].sharex(axes[2, 1])
-    plt.setp(axes[2, 1].get_xticklabels(), visible=False)
-    axes[3, 1].set_xlim([-2, 2])
-
-    exp_data = simulations.data.load_SiO_x_multistate()
-    V, I = simulations.data.all_SiO_x_curves(exp_data, clean_data=True)
-    resistances, c, d_times_perm, _, _ = simulations.data.pf_relationship(V, I)
-    R_0 = const.physical_constants["inverse of conductance quantum"][0]
-    # Separate data into before and after the conductance quantum.
-    sep_idx = np.searchsorted(resistances, R_0)
-
-    colors = utils.color_dict()
-
-    axes[0, 0].set_ylabel(utils.axis_label("ln-c-residuals"))
-    axes[1, 0].set_ylabel(utils.axis_label("ln-d-times-perm-residuals"))
-    axes[1, 0].set_xlabel(utils.axis_label("ln-R"))
-    axes[1, 1].set_xlabel(utils.axis_label("ln-R"))
-    axes[3, 0].set_xlabel(utils.axis_label("theoretical-normal-quartiles"))
-    axes[2, 0].set_ylabel(utils.axis_label("ordered-ln-c-residuals"))
-    axes[3, 1].set_xlabel(utils.axis_label("theoretical-normal-quartiles"))
-    axes[3, 0].set_ylabel(utils.axis_label("ordered-ln-d-times-perm-residuals"))
-
-    for is_high_resistance, ax_idx in zip([False, True], [0, 1]):
-        _, _, slopes, intercepts, _ = simulations.data.pf_params(
-            exp_data, is_high_resistance, simulations.data.SiO_x_G_on_G_off_ratio()
-        )
-
-        if is_high_resistance:
-            idxs = np.arange(sep_idx, len(resistances))
-            color = colors["reddish-purple"]
-        else:
-            idxs = np.arange(sep_idx)
-            color = colors["blue"]
-
-        x = np.log(resistances[idxs])
-        c_points = np.log(c[idxs])
-        c_residuals = c_points - slopes[0] * x - intercepts[0]
-
-        d_times_perm_points = np.log(d_times_perm[idxs])
-        d_times_perm_residuals = d_times_perm_points - slopes[1] * x - intercepts[1]
-
-        utils.plot_scatter(axes[0, ax_idx], x, c_residuals, color, scale=10)
-        zero_line = [0] * len(x)
-        axes[0, ax_idx].plot(
-            x,
-            zero_line,
-            linewidth=utils.Config.LINEWIDTH,
-            color=color,
-            linestyle="dashed",
-        )
-
-        utils.plot_scatter(axes[1, ax_idx], x, d_times_perm_residuals, color, scale=10)
-        axes[1, ax_idx].plot(
-            x,
-            zero_line,
-            linewidth=utils.Config.LINEWIDTH,
-            color=color,
-            linestyle="dashed",
-        )
-
-        (osm, osr), (m, b, _) = stats.probplot(c_residuals, dist="norm", plot=None)
-        utils.plot_scatter(axes[2, ax_idx], osm, osr, color, scale=10)
-        axes[2, ax_idx].plot(
-            osm,
-            m * osm + b,
-            linewidth=utils.Config.LINEWIDTH,
-            color=color,
-            linestyle="dashed",
-        )
-
-        (osm, osr), (m, b, _) = stats.probplot(d_times_perm_residuals, dist="norm", plot=None)
-        utils.plot_scatter(axes[3, ax_idx], osm, osr, color, scale=10)
-        axes[3, ax_idx].plot(
-            osm,
-            m * osm + b,
-            linewidth=utils.Config.LINEWIDTH,
-            color=color,
-            linestyle="dashed",
-        )
-
-    for ax in [axes[0, 0], axes[1, 0]]:
-        low, high = ax.get_ylim()
-        bound = max(abs(low), abs(high))
-        ax.set_ylim(-bound, bound)
-
-    utils.save_fig(fig, "pf-residuals")
+    utils.save_fig(fig, "Ta-HfO2")
 
 
 def iv_nonlinearity_training(metric="error"):
@@ -685,3 +581,107 @@ def nonideality_agnosticism(metric: str = "error", norm_rows=True, include_val_l
         )
 
     utils.save_fig(fig, filename, metric=metric)
+
+
+def pf_residuals(is_d_times_perm: bool = False):
+    fig, axes = utils.fig_init(2, 1.0, fig_shape=(4, 2))
+
+    axes[0, 1].sharey(axes[0, 0])
+    plt.setp(axes[0, 1].get_yticklabels(), visible=False)
+    axes[1, 1].sharey(axes[1, 0])
+    plt.setp(axes[1, 1].get_yticklabels(), visible=False)
+
+    axes[1, 0].sharex(axes[0, 0])
+    plt.setp(axes[0, 0].get_xticklabels(), visible=False)
+    axes[1, 1].sharex(axes[0, 1])
+    plt.setp(axes[0, 1].get_xticklabels(), visible=False)
+
+    axes[3, 0].sharex(axes[2, 0])
+    plt.setp(axes[2, 0].get_xticklabels(), visible=False)
+    axes[3, 0].set_xlim([-2, 2])
+    axes[3, 1].sharex(axes[2, 1])
+    plt.setp(axes[2, 1].get_xticklabels(), visible=False)
+    axes[3, 1].set_xlim([-2, 2])
+
+    exp_data = simulations.data.load_SiO_x_multistate()
+    V, I = simulations.data.all_SiO_x_curves(exp_data, clean_data=True)
+    resistances, c, d_times_perm, _, _ = simulations.data.pf_relationship(V, I)
+    R_0 = const.physical_constants["inverse of conductance quantum"][0]
+    # Separate data into before and after the conductance quantum.
+    sep_idx = np.searchsorted(resistances, R_0)
+
+    colors = utils.color_dict()
+
+    axes[0, 0].set_ylabel(utils.axis_label("ln-c-residuals"))
+    axes[1, 0].set_ylabel(utils.axis_label("ln-d-times-perm-residuals"))
+    axes[1, 0].set_xlabel(utils.axis_label("ln-R"))
+    axes[1, 1].set_xlabel(utils.axis_label("ln-R"))
+    axes[3, 0].set_xlabel(utils.axis_label("theoretical-normal-quartiles"))
+    axes[2, 0].set_ylabel(utils.axis_label("ordered-ln-c-residuals"))
+    axes[3, 1].set_xlabel(utils.axis_label("theoretical-normal-quartiles"))
+    axes[3, 0].set_ylabel(utils.axis_label("ordered-ln-d-times-perm-residuals"))
+
+    for is_high_resistance, ax_idx in zip([False, True], [0, 1]):
+        _, _, slopes, intercepts, _ = simulations.data.pf_params(
+            exp_data, is_high_resistance, simulations.data.SiO_x_G_on_G_off_ratio()
+        )
+
+        if is_high_resistance:
+            idxs = np.arange(sep_idx, len(resistances))
+            color = colors["reddish-purple"]
+        else:
+            idxs = np.arange(sep_idx)
+            color = colors["blue"]
+
+        x = np.log(resistances[idxs])
+        c_points = np.log(c[idxs])
+        c_residuals = c_points - slopes[0] * x - intercepts[0]
+
+        d_times_perm_points = np.log(d_times_perm[idxs])
+        d_times_perm_residuals = d_times_perm_points - slopes[1] * x - intercepts[1]
+
+        utils.plot_scatter(axes[0, ax_idx], x, c_residuals, color, scale=10)
+        zero_line = [0] * len(x)
+        axes[0, ax_idx].plot(
+            x,
+            zero_line,
+            linewidth=utils.Config.LINEWIDTH,
+            color=color,
+            linestyle="dashed",
+        )
+
+        utils.plot_scatter(axes[1, ax_idx], x, d_times_perm_residuals, color, scale=10)
+        axes[1, ax_idx].plot(
+            x,
+            zero_line,
+            linewidth=utils.Config.LINEWIDTH,
+            color=color,
+            linestyle="dashed",
+        )
+
+        (osm, osr), (m, b, _) = stats.probplot(c_residuals, dist="norm", plot=None)
+        utils.plot_scatter(axes[2, ax_idx], osm, osr, color, scale=10)
+        axes[2, ax_idx].plot(
+            osm,
+            m * osm + b,
+            linewidth=utils.Config.LINEWIDTH,
+            color=color,
+            linestyle="dashed",
+        )
+
+        (osm, osr), (m, b, _) = stats.probplot(d_times_perm_residuals, dist="norm", plot=None)
+        utils.plot_scatter(axes[3, ax_idx], osm, osr, color, scale=10)
+        axes[3, ax_idx].plot(
+            osm,
+            m * osm + b,
+            linewidth=utils.Config.LINEWIDTH,
+            color=color,
+            linestyle="dashed",
+        )
+
+    for ax in [axes[0, 0], axes[1, 0]]:
+        low, high = ax.get_ylim()
+        bound = max(abs(low), abs(high))
+        ax.set_ylim(-bound, bound)
+
+    utils.save_fig(fig, "pf-residuals")

awarememristor/simulations/ideal.py

@@ -1,4 +1,4 @@
-from awarememristor.simulations import (iv_nonlinearity,
+from awarememristor.simulations import (high_d2d, iv_nonlinearity,
                                         iv_nonlinearity_and_stuck_on,
                                         iv_nonlinearity_cnn,
                                         stuck_distribution, stuck_off,
@@ -13,7 +13,7 @@ def get_mnist_iterator():
         iv_nonlinearity_and_stuck_on.get_ideal_iterator(),
         stuck_distribution.get_ideal_iterator(),
         weight_implementation.get_ideal_iterator(),
-        memristive_validation.get_ideal_iterator(),
+        high_d2d.get_ideal_iterator(),
     ]
 
     return Iterator(

reproduce_paper.py

@@ -15,28 +15,30 @@
 simulations.iv_nonlinearity.main()
 simulations.iv_nonlinearity_and_stuck_on.main()
 simulations.iv_nonlinearity_cnn.main()
-simulations.memristive_validation.main()
 simulations.stuck_distribution.main()
 simulations.stuck_off.main()
+simulations.high_d2d.main()
 simulations.weight_implementation.main()
 simulations.nonideality_agnosticism.main()
 
 # Figures in the main text
-figures.experimental_data()
+figures.SiO_x()
+figures.Ta_HfO2()
 figures.iv_nonlinearity_training()
 figures.iv_nonlinearity_inference()
 figures.iv_nonlinearity_cnn()
 figures.weight_implementation()
-figures.memristive_validation()
 figures.nonideality_agnosticism()
+figures.pf_residuals()
 
 # Figures in the Supporting Information
 supporting_figures.all_iv_curves_full_range()
 supporting_figures.switching()
 supporting_figures.iv_nonlinearity_training()
 supporting_figures.weight_implementation_standard_weights_training()
 supporting_figures.weight_implementation_double_weights_training()
-supporting_figures.memristive_validation_training()
+supporting_figures.high_d2d_training()
 supporting_figures.stuck_off_training()
 supporting_figures.high_iv_nonlinearity_and_stuck_on_training()
 supporting_figures.stuck_distribution_training()
+supporting_figures.pf_plots()