SciML

Scientific Machine Learning

Author: dimtsap

README.rst

@@ -35,7 +35,7 @@ Uncertainty Quantification with python (UQpy)
      - Aakash Bangalore Satish, Mohit Singh Chauhan, Lohit Vandanapu, Ketson RM dos Santos, Katiana Kontolati, Dimitris Loukrezis, Promit Chakroborty, Lukáš Novák, Andrew Solanto, Connor Krill
 
    * - **Contributors:**
-     - Michael Gardner, Prateek Bhustali, Julius Schultz, Ulrich Römer
+     - Michael Gardner, Prateek Bhustali, Julius Schultz, Ulrich Römer, Nicholas Betters
 
 
 

docs/code/scientific_machine_learning/bayesian_quickstart/bayesian_quickstart_testing.py

@@ -8,15 +8,15 @@
 # https://pytorch.org/tutorials/beginner/basics/quickstart_tutorial.html
 #
 # This script assumes you have already run the Bayesian Quickstart Testing script and saved the optimized model
-# to a file named ``bayesian_model.pth``. This script
+# to a file named ``bayesian_model.pt``. This script
 #
 # - Loads a trained model from the file ``bayesian_model.pt``
 # - Makes deterministic predictions
 # - Makes probabilistic predictions
 # - Plots probabilistic predictions
 #
 # First, we import the necessary modules and define the BayesianNeuralNetwork class, so we can load the model state
-# dictionary saved in ``bayesian_model.pth``.
+# dictionary saved in ``bayesian_model.pt``.
 
 # %%
 import torch

docs/code/scientific_machine_learning/bayesian_quickstart/bayesian_quickstart_training.py

@@ -5,7 +5,7 @@
 
 # %% md
 # This is the first half of a Bayesian version of the classification problem from this Pytorch Quickstart tutorial:
-# (https://pytorch.org/tutorials/beginner/basics/quickstart_tutorial.html)
+# https://pytorch.org/tutorials/beginner/basics/quickstart_tutorial.html
 #
 # We strongly recommend reading the Pytorch quick start first to familiarize yourself with the problem.
 # We include many of the comments from the Pytorch example, but assume the reader is familiar with model definitions

docs/code/scientific_machine_learning/mcd_trainer/NeuralNetwork_MCD.py

@@ -37,7 +37,7 @@ def __init__(self, n_samples=20, noise_std=0.05):
         self.n_samples = n_samples
         self.noise_std = noise_std
         self.x = torch.linspace(-1, 1, n_samples).reshape(-1, 1)
-        self.y =0.4 * torch.sin(4 * self.x) + 0.5 * torch.cos(12 * self.x)
+        self.y = 0.4 * torch.sin(4 * self.x) + 0.5 * torch.cos(12 * self.x)
         self.y += torch.normal(0, self.noise_std, self.x.shape)
 
     def __len__(self):
@@ -147,78 +147,4 @@ def __getitem__(self, item):
 ax.set(xlabel="x", ylabel="f(x)")
 ax.legend()
 
-
-# Plotting Results
-# x_data = SinusoidalDataset().x.detach()
-# y_data = SinusoidalDataset().y.detach()
-# x_val = torch.linspace(-1, 1, 1000).view(-1, 1).detach()
-# y_val = 0.4 * torch.sin(4 * x_val) + 0.5 * torch.cos(12 * x_val).detach()
-# pred_val = model(x_val).detach()
-#
-# # %% Plot the deterministic model estimates
-# fig, ax = plt.subplots()
-# ax.scatter(x_data, y_data, label="Data", color="black", s=50)
-# ax.plot(
-#     x_val,
-#     pred_val,
-#     label="Final Prediction",
-#     color="tab:orange",
-# )
-# ax.plot(
-#     x_val.detach(),
-#     y_val.detach(),
-#     label="Target",
-#     color="black",
-#     linestyle="dashed",
-# )
-# ax.set_title("Deterministic Prediction")
-# ax.set(xlabel="$x$", ylabel="$f(x)$")
-# ax.legend()
-# fig.tight_layout()
-#
-# train_loss = trainer.history["train_loss"].detach().numpy()
-# fig, ax = plt.subplots()
-# ax.semilogy(train_loss)
-# ax.set_title("Training Loss")
-# ax.set(xlabel="Epoch", ylabel="Loss")
-# fig.tight_layout()
-#
-# # %%
-# model.drop()  # activate the dropout layers
-# n = 1_000
-# samples = torch.zeros(n, len(x_val))
-# for i in range(n):
-#     samples[i, :] = model(x_val).detach().squeeze()
-# mean = torch.mean(samples, dim=1)
-# standard_deviation = torch.std(samples, dim=1)
-#
-# # Plotting Results
-# fig, ax = plt.subplots()
-# ax.plot(x_val, samples[:, 1], "tab:orange", label="Prediction 1")
-# ax.plot(x_val, samples[:, 2], "tab:blue", label="Prediction 2")
-# ax.scatter(x_data, y_data, color="black", label="Data")
-# ax.plot(x_val, y_val, color="black", linestyle="dashed", label="Target")
-# ax.set_title("Two Samples from Dropout NN")
-# ax.set(xlabel="$x$", ylabel="$f(x)$")
-# ax.legend()
-# fig.tight_layout()
-#
-# # %%
-# fig, ax = plt.subplots()
-# ax.plot(x_val, mean, label="$\mu$")
-# ax.fill_between(
-#     x_val.view(-1),
-#     torch.quantile(samples, q=0.025, dim=1),
-#     torch.quantile(samples, q=0.975, dim=1),
-#     label="95% Range",
-#     # mean - (3 * standard_deviation),
-#     # mean + (3 * standard_deviation),
-#     # label="$\mu \pm 3\sigma$,",
-#     alpha=0.3,
-# )
-# ax.plot(x_val, y_val, label="Target", color="black")
-# ax.set_title("Dropout Neural Network 95% Range")
-# ax.set(xlabel="x", ylabel="f(x)")
-# ax.legend()
-
 plt.show()

docs/requirements.txt

@@ -3,4 +3,6 @@ sphinx_autodoc_typehints == 1.23.0
 sphinx_rtd_theme == 1.2.0
 sphinx_gallery == 0.13.0
 sphinxcontrib_bibtex == 2.5.0
-Sphinx==6.1.3
+Sphinx==6.1.3
+torch == 2.6.0
+torchinfo ~= 1.8.0

docs/source/conf.py

@@ -25,7 +25,7 @@
 )
 
 # The full version, including alpha/beta/rc tags
-release = "v4.1.7"
+release = "v4.2.0"
 
 # -- General configuration ---------------------------------------------------
 

docs/source/index.rst

@@ -12,15 +12,21 @@ as a set of modules centered around core capabilities in Uncertainty Quantificat
 +-----------------------+------------------------------------------------------------------+
 | **Product Owner:**    | Michael D. Shields                                               |
 +-----------------------+------------------------------------------------------------------+
-| **Lead Developers:**  | Dimitris Giovanis, Audrey Olivier, Dimitris Tsapetis             |
+| **Lead Developers:**  | Dimitris Giovanis, Audrey Olivier, Dimitris Tsapetis,            |
++                       +                                                                  +
+|                       | Connor Krill                                                     |
 +-----------------------+------------------------------------------------------------------+
 | **Development Team:** | Aakash Bangalore Satish, Mohit Singh Chauhan, Lohit Vandanapu,   |
 +                       +                                                                  +
 |                       | Ketson RM dos Santos, Katiana Kontolati, Dimitris Loukrezis,     |
 +                       +                                                                  +
-|                       | Promit Chakroborty, Lukáš Novák, Andrew Solanto, Connor Krill    |
+|                       | Promit Chakroborty, Lukáš Novák, Andrew Solanto,                 |
++                       +                                                                  +
+|                       | Ponkrshnan Thiagarajan, George Pasparakis                        |
 +-----------------------+------------------------------------------------------------------+
-| **Contributors:**     | Michael Gardner, Prateek Bhustali, Julius Schultz, Ulrich Römer  |
+| **Contributors:**     | Michael Gardner, Prateek Bhustali, Julius Schultz, Ulrich Römer, |
++                       +                                                                  +
+|                       | Nicholas Betters                                                 |
 +-----------------------+------------------------------------------------------------------+
 
 Introduction

docs/source/scientific_machine_learning/index.rst

@@ -20,8 +20,8 @@ which uses similar inputs.
 
 The module contains the following parent classes for neural networks:
 
-- :class:`.NormalBayesianLayer`: Parent class to all Bayesian layers. Subclass of :class:`Layer`
-- :class:`.ProbabilisticDropoutLayer`: Parent class to all Dropout layers. Subclass of :class:`Layer`
+- :class:`.NormalBayesianLayer`: Parent class to all Bayesian layers. Subclass of :class:`Layer`.
+- :class:`.ProbabilisticDropoutLayer`: Parent class to all Dropout layers. Subclass of :class:`Layer`.
 - :class:`.Layer`: Parent class to all Neural Network Layers. Subclass of :class:`torch.nn.Module`.
 - :class:`.Loss`: Parent class to all Loss functions. Subclass of :class:`torch.nn.Module`.
 - :class:`.NeuralNetwork`: Parent class to all Neural Networks and Neural Operators.  Subclass of :class:`torch.nn.Module`.

docs/source/scientific_machine_learning/layers/index.rst

@@ -42,7 +42,7 @@ This is the parent class to all layers.
 Like all abstract baseclasses, this cannot be instantiated but can be subclassed to write custom layers.
 All layers use the :py:meth:`forward` method to define the forward model call.
 
-The documentation in the :py:meth:`forward` and :py:meth:`extra_repr` on this page may be inherited from PyTorch docstrings.
+Some documentation within the :class:`Layer` class may be inherited from PyTorch docstrings.
 
 Methods
 ~~~~~~~

docs/source/scientific_machine_learning/losses.rst

@@ -15,7 +15,7 @@ The :py:class:`Loss` is an abstract baseclass and a subclass of :py:class:`torch
 This is an abstract baseclass and the parent class to all loss functions.
 Like all abstract baseclasses, this cannot be instantiated but can be subclassed to write custom losses.
 
-The documentation in the :py:meth:`forward` on this baseclass may be inherited from PyTorch docstrings.
+The documentation in the :class:`Loss` may be inherited from PyTorch docstrings.
 
 Methods
 ~~~~~~~