Adds vectorized time integration to pyoptmat (#23)

* Adds functionality for sparse matrices/sparse solvers * Adds specialized bidiagonal sparse matrices/solvers. * Implements chunked time integration * Updates time/temperature interpolation routines to better take advantage of the new functionality. --------- Co-authored-by: Tianju Chen <[email protected]>
Argonne-National-Laboratory · Mar 31, 2023 · 6b17430 · 6b17430
1 parent d0a8d25
commit 6b17430
Show file tree

Hide file tree

Showing 54 changed files with 3,403 additions and 1,984 deletions.
diff --git a/.gitignore b/.gitignore
@@ -19,3 +19,6 @@ doc/sphinx/_build
 build/
 dist/
 pyoptmat.egg-info/
+examples/ode/results.txt
+examples/ode/log
+results.txt
diff --git a/doc/sphinx/chunktime.rst b/doc/sphinx/chunktime.rst
@@ -0,0 +1,7 @@
+pyoptmat.chunktime: utilities for blocked time integration
+----------------------------------------------------------
+
+.. automodule:: pyoptmat.chunktime
+   :members:
+   :undoc-members:
+   :show-inheritance:
diff --git a/doc/sphinx/conf.py b/doc/sphinx/conf.py
@@ -22,7 +22,7 @@
 author = 'Argonne National Laboratory'
 
 # The full version, including alpha/beta/rc tags
-release = '1.1.3'
+release = '1.2.0'
 
 
 # -- General configuration ---------------------------------------------------

diff --git a/doc/sphinx/examples.rst b/doc/sphinx/examples.rst
@@ -28,6 +28,14 @@ repeats the :doc:`tutorial <structural_tutorial>`, but using
 synthetic cyclic test data a :py:class:`pyoptmat.hardening.ChabocheHardeningModel` viscoplastic model capable
 of capturing kinematic hardening behavior.
 
+Neural ODE example
+------------------
+
+Inferring the response of a coupled mass-spring-damper system using
+trajectory data (1) a physical ODE model and (2) a neural ODE model.
+
+.. literalinclude:: /../../examples/ode/damping.py
+
 Implicit versus explicit integration
 ------------------------------------
 
@@ -57,6 +65,13 @@ performance of both approaches.
 
 .. literalinclude:: /../../examples/ode/trajectory.py
 
+Hogdkin-Huxley coupled neuron model
+-----------------------------------
+
+An example, scalable system of ODEs for benchmarking.
+
+.. literalinclude:: /../../examples/ode/neuron.py
+
 Temperature dependent parameters
 --------------------------------
 

diff --git a/doc/sphinx/index.rst b/doc/sphinx/index.rst
@@ -40,6 +40,8 @@ pyoptmat features
   uncertain data.
 - Efficient backward pass/gradient calculation using the adjoint method.  This
   approach vastly outperforms automatic differentiation for time series data. 
+- Blocked time integration for both the forward and backward/adjoint passes, which 
+  vectorizes/parallelizes integrating ODEs through time.
 - Implicit time integration algorithms suitable for material models represents
   as stiff systems of ordinary differential equations.
 - Prebuilt model components aimed at high temperature structural materials.
@@ -78,6 +80,7 @@ The following are links to the complete API descriptions of each pyoptmat submod
    optimize
    experiments
    ode
+   chunktime
    solvers
    models
    flowrules

diff --git a/doc/sphinx/refs.bib b/doc/sphinx/refs.bib
@@ -38,3 +38,12 @@ @article{chaboche1989unified
   volume={111},
   pages={424--430}
 }
+
+@article{schwemmer2012theory,
+  title={The theory of weakly coupled oscillators},
+  author={Schwemmer, Michael A and Lewis, Timothy J},
+  journal={Phase response curves in neuroscience: Theory, experiment, and analysis},
+  pages={3--31},
+  year={2012},
+  publisher={Springer}
+}