Sebastian.p/orth dgp

CONTRIBUTORS.md

@@ -12,7 +12,9 @@ Because GitHub's [graph of contributors](http://github.com/secondmind-labs/GPflu
 [Felix Leibfried](https://github.com/fleibfried), 
 [John A. McLeod](https://github.com/johnamcleod), 
 [Hugh Salimbeni](https://github.com/hughsalimbeni), 
-[Marcin B. Tomczak](https://github.com/marctom)
+[Marcin B. Tomczak](https://github.com/marctom),
+[Sebastian Popescu](https://github.com/SebastianPopescu),
+[Alessandro Vullo](https://github.com/avullo),
 
 
 Feel free to add yourself when you first contribute to GPflux's code, tests, or documentation!

docs/notebooks/gpflux_features.py

@@ -61,15 +61,22 @@ def motorcycle_data():
 """
 
 # %%
-import gpflux
+from gpflow.kernels import SquaredExponential
 
-from gpflux.architectures import Config, build_constant_input_dim_deep_gp
+import gpflux
+from gpflux.architectures.config import GaussianLikelihoodConfig, ModelHyperParametersConfig
+from gpflux.architectures.factory import build_constant_input_dim_architecture
 from gpflux.models import DeepGP
 
-config = Config(
-    num_inducing=25, inner_layer_qsqrt_factor=1e-5, likelihood_noise_variance=1e-2, whiten=True
+config = ModelHyperParametersConfig(
+    num_layers=2,
+    kernel=SquaredExponential,
+    likelihood=GaussianLikelihoodConfig(noise_variance=1e-2),
+    inner_layer_qsqrt_factor=1e-5,
+    whiten=True,
+    num_inducing=25,
 )
-deep_gp: DeepGP = build_constant_input_dim_deep_gp(X, num_layers=2, config=config)
+deep_gp: DeepGP = build_constant_input_dim_architecture(config, X)
 
 # %% [markdown]
 """
@@ -164,9 +171,7 @@ def plot(model, X, Y, ax=None):
 prediction_model.save_weights("weights")
 
 # %%
-prediction_model_new = build_constant_input_dim_deep_gp(
-    X, num_layers=2, config=config
-).as_prediction_model()
+prediction_model_new = build_constant_input_dim_architecture(config, X).as_prediction_model()
 prediction_model_new.load_weights("weights")
 
 # %%

docs/notebooks/plotting_functions.py

@@ -0,0 +1,195 @@
+#
+# Copyright (c) 2022 The GPflux Contributors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+"""
+This module contains experimental code for plotting. As such, the disclaimer is clear.
+"""
+import numpy as np
+import matplotlib.pyplot as plt
+import tensorflow as tf
+import io
+
+
+def get_classification_detailed_plot(
+    num_layers, X_training, Y_training, where_to_save, f_mean_overall, f_var_overall, name_file
+):
+
+    xx, yy = np.mgrid[-5:5:0.1, -5:5:0.1]
+    grid = np.c_[xx.ravel(), yy.ravel()]
+    grid = grid.astype(np.float32)
+
+    indices_class_1 = np.where(Y_training == 1.0)
+    indices_class_0 = np.where(Y_training == 0.0)
+
+    fig, axs = plt.subplots(
+        nrows=2, ncols=num_layers, sharex=True, sharey=True, figsize=(20 * num_layers, 40)
+    )
+
+    for current_layer in range(num_layers):
+
+        current_mean = f_mean_overall[current_layer]
+        current_mean = current_mean.reshape((100, 100))
+        current_var = f_var_overall[current_layer]
+        current_var = current_var.reshape((100, 100))
+
+        ###################
+        ##### F mean  #####
+        ###################
+
+        axis = axs[0, current_layer]
+        contour = axis.contourf(xx, yy, current_mean, 50, cmap="coolwarm")
+        cbar1 = fig.colorbar(contour, ax=axis)
+
+        cbar1.ax.tick_params(labelsize=60)
+
+        axis.set(xlim=(-5.0, 5.0), ylim=(-5.0, 5.0), xlabel="$X_1$", ylabel="$X_2$")
+        axis.set_title(label="Predictive Mean", fontdict={"fontsize": 60})
+        axis.tick_params(axis="both", which="major", labelsize=80)
+
+        axis.scatter(
+            X_training[indices_class_0, 0],
+            X_training[indices_class_0, 1],
+            s=100,
+            marker="X",
+            alpha=0.2,
+            c="green",
+            linewidth=1,
+            label="Class 0",
+        )
+        axis.scatter(
+            X_training[indices_class_1, 0],
+            X_training[indices_class_1, 1],
+            s=100,
+            marker="D",
+            alpha=0.2,
+            c="purple",
+            linewidth=1,
+            label="Class 1",
+        )
+
+        # NOTE -- don't need this
+        # axis.scatter(Z_np[current_layer][:,0], Z_np[current_layer][:,1],
+        #        s=750, marker="*", alpha=0.95, c = 'cyan',
+        #        linewidth=1, label = 'Inducing Points')
+
+        axis.legend(loc="upper right", prop={"size": 60})
+        # axis.text(-4.5, 4.5, 'LL:'+"{:.2f}".format(total_nll_np)+'; Acc:'+"{:.2f}".format(precision_testing_overall_np), size=50, color='black')
+
+        #################################
+        ##### F var Distributional  #####
+        #################################
+
+        axis = axs[1, current_layer]
+        contour = axis.contourf(xx, yy, current_var, 50, cmap="coolwarm")
+        cbar1 = fig.colorbar(contour, ax=axis)
+        cbar1.ax.tick_params(labelsize=60)
+
+        axis.set(xlim=(-5, 5), ylim=(-5, 5), xlabel="$X_1$", ylabel="$X_2$")
+
+        axis.set_title(label="Predictive Variance", fontdict={"fontsize": 60})
+        axis.tick_params(axis="both", which="major", labelsize=80)
+
+        axis.scatter(
+            X_training[indices_class_0, 0],
+            X_training[indices_class_0, 1],
+            s=100,
+            marker="X",
+            alpha=0.2,
+            c="green",
+            linewidth=1,
+            label="Class 0",
+        )
+        axis.scatter(
+            X_training[indices_class_1, 0],
+            X_training[indices_class_1, 1],
+            s=100,
+            marker="D",
+            alpha=0.2,
+            c="purple",
+            linewidth=1,
+            label="Class 1",
+        )
+
+        # axis.scatter(Z_np[current_layer][:,0], Z_np[current_layer][:,1],
+        #    s=750, marker="*", alpha=0.95, c = 'cyan',
+        #    linewidth=1, label = 'Inducing Points')
+        axis.legend(loc="upper right", prop={"size": 60})
+
+    plt.tight_layout()
+    plt.savefig(where_to_save + name_file)
+    plt.close()
+
+
+def get_regression_detailed_plot(
+    num_layers,
+    X_training,
+    Y_training,
+    where_to_save,
+    mean,
+    var,
+    name_file,
+    x_margin,
+    y_margin,
+    X_test,
+):
+
+    figure, axs = plt.subplots(
+        nrows=1, ncols=num_layers, sharex=True, sharey=True, figsize=(10 * num_layers, 10)
+    )
+
+    for current_layer in range(num_layers):
+        current_mean = mean[current_layer]
+        current_var = var[current_layer]
+
+        ###################
+        ##### F mean  #####
+        ###################
+
+        X_test = X_test.squeeze()
+        lower = current_mean - 2 * np.sqrt(current_var)
+        upper = current_mean + 2 * np.sqrt(current_var)
+
+        axis = axs[current_layer]
+        axis.set_ylim(Y_training.min() - y_margin, Y_training.max() + y_margin)
+        axis.plot(X_training, Y_training, "kx", alpha=0.5, label="Training")
+        axis.plot(X_test, current_mean, "C1")
+
+        axis.fill_between(X_test, lower, upper, color="C1", alpha=0.3)
+        axis.legend(loc="upper right", prop={"size": 60})
+
+        axis.set_title(label=f"Layer {current_layer+1}", fontdict={"fontsize": 60})
+        axis.tick_params(axis="both", which="major", labelsize=80)
+
+    plt.tight_layout()
+    # plt.savefig(where_to_save+name_file)
+    # plt.close()
+    return figure
+
+
+def plot_to_image(figure):
+    """Converts the matplotlib plot specified by 'figure' to a PNG image and
+    returns it. The supplied figure is closed and inaccessible after this call."""
+    # Save the plot to a PNG in memory.
+    buf = io.BytesIO()
+    plt.savefig(buf, format="png")
+    # Closing the figure prevents it from being displayed directly inside
+    # the notebook.
+    plt.close(figure)
+    buf.seek(0)
+    # Convert PNG buffer to TF image
+    image = tf.image.decode_png(buf.getvalue(), channels=4)
+    # Add the batch dimension
+    image = tf.expand_dims(image, 0)
+    return image