Fides 0.6.1 (#39)

* fix nan values in hessian

* version bump

* add reference, fixes #36

* fix flake

* update for 3.10

* fix 3.10

* update readme

* wait with 3.10 until scipy gets it shit together

* don't bump requirement just yet

.github/workflows/tests.yaml

@@ -9,7 +9,7 @@ jobs:
 
     strategy:
       matrix:
-        python-version: [3.7, 3.8, 3.9]
+        python-version: ['3.7', '3.8', '3.9']
 
     steps:
 

CITATION.cff

@@ -0,0 +1,28 @@
+cff-version: 1.2.0
+message: "If you use this software, please cite it as below."
+authors:
+- family-names: "Froehlich"
+  given-names: "Fabian"
+  orcid: "https://orcid.org/0000-0003-2387-9495"
+- family-names: "Weindl"
+  given-names: "Daniel"
+  orcid: "https://orcid.org/0000-0001-9963-6057"
+title: "Fides"
+version: 0.6.0
+doi: 10.5281/zenodo.5554996
+date-released: 2021-10-07
+url: "https://github.com/fides/fides"
+preferred-citation:
+  type: article
+  authors:
+  - family-names: "Froehlich"
+    given-names: "Fabian"
+    orcid: "https://orcid.org/0000-0003-2387-9495"
+  - family-names: "Sorger"
+    given-names: "Peter K."
+    orcid: "https://orcid.org/0000-0002-3364-1838"
+  doi: "10.1101/2021.05.20.4450650"
+  journal: "bioRxiv"
+  month: 5
+  title: "Fides: Reliable Trust-Region Optimization for Parameter Estimation of Ordinary Differential Equation Models"
+  year: 2021

README.md

@@ -25,8 +25,10 @@ Fides can be installed via `pip install fides`. Further documentation is
 ## Features
 
 
-* Boundary constrained interior trust-region optimization
+* Boundary constrained and unconstrained interior trust-region optimization
 * Reflective, truncated and optimization based boundary heuristics
 * Exact, 2D and CG subproblem solvers
-* BFGS, DFP, SR1 and Hybrid  Hessian Approximations
+* BFGS, DFP, SR1, PSB, Broyden (good and bad) and Broyden class iterative
+ Hessian Approximation schemes
+* SSM, TSSM, FX, GNSBFGS and custom hybrid Hessian Approximations schemes
 

fides/hessian_approximation.py

@@ -440,7 +440,7 @@ def broyden_class_update(y, s, mat, phi=None, v=None):
         return np.outer(y, y.T) / b - np.outer(u, u.T) / c
 
     if v is None:
-        rho = np.sqrt(b / c)
+        rho = np.sqrt(b / c) if c > 0 else 0  # c == 0 iff u == 0
         v = y + (1-phi) * rho * u
 
     z = y - mat.dot(s)

fides/minimize.py

@@ -294,6 +294,8 @@ def minimize(self, x0: np.ndarray):
             theta = max(self.get_option(Options.THETA_MAX),
                         1 - norm(v * self.grad, np.inf))
 
+            self.check_finite()
+
             step = \
                 trust_region(
                     self.x, self.grad, self.hess, scaling,
@@ -667,7 +669,7 @@ def log_header(self):
             f'| step | refl | trun | accept'
         )
 
-    def check_finite(self, funout: Funout):
+    def check_finite(self, funout: Optional[Funout] = None):
         """
         Checks whether objective function value, gradient and Hessian (
         approximation) have finite values and optimization can continue.
@@ -684,7 +686,10 @@ def check_finite(self, funout: Funout):
         else:
             pointstr = f'at iteration {self.iteration}.'
 
-        fval, grad, hess = funout.fval, funout.grad, funout.hess
+        if funout is not None:
+            fval, grad, hess = funout.fval, funout.grad, funout.hess
+        else:
+            fval, grad, hess = self.fval, self.grad, self.hess
 
         if not np.isfinite(fval):
             self.exitflag = ExitFlag.NOT_FINITE

fides/steps.py

@@ -284,9 +284,9 @@ def __init__(self, x, sg, hess, scaling, g_dscaling, delta, theta,
 
 class CGStep(Step):
     """
-        This class provides the machinery to compute an approximate solution of
-        the trust region subproblem using the Steihaug Method
-        """
+    This class provides the machinery to compute an approximate solution of
+    the trust region subproblem using the conjugate gradients methods
+    """
 
     type = 'cg'
 
@@ -310,7 +310,7 @@ def conj_grad(self, eps):
 class TRStepSteihaug(CGStep):
     """
     This class provides the machinery to compute an approximate solution of
-    the trust region subproblem using the Steihaug Method
+    the trust region subproblem using Steihaug's Method
     """
 
     type = 'cgs'

fides/version.py

@@ -1 +1 @@
-__version__ = "0.6.0"
+__version__ = "0.6.1"