diffaaable 1.0.1__tar.gz → 1.1.1__tar.gz

{diffaaable-1.0.1 → diffaaable-1.1.1}/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.1
+Metadata-Version: 2.3
 Name: diffaaable
-Version: 1.0.1
+Version: 1.1.1
 Summary: JAX-differentiable AAA algorithm
 Keywords: python
 Author-email: Jan David Fischbach <fischbach@kit.edu>
@@ -15,12 +15,14 @@ Requires-Dist: jax
 Requires-Dist: jaxlib
 Requires-Dist: baryrat
 Requires-Dist: jaxopt
-Requires-Dist: pytest
-Requires-Dist: pytest-benchmark
+Requires-Dist: tbump>=6.11.0 ; extra == "dev"
+Requires-Dist: towncrier ; extra == "dev"
 Requires-Dist: pre-commit ; extra == "dev"
 Requires-Dist: pytest ; extra == "dev"
 Requires-Dist: pytest-cov ; extra == "dev"
 Requires-Dist: pytest_regressions ; extra == "dev"
+Requires-Dist: pytest-benchmark ; extra == "dev"
+Requires-Dist: matplotlib ; extra == "dev"
 Requires-Dist: jupytext ; extra == "docs"
 Requires-Dist: matplotlib ; extra == "docs"
 Requires-Dist: jupyter-book ; extra == "docs"
@@ -28,7 +30,7 @@ Requires-Dist: sphinx_math_dollar ; extra == "docs"
 Provides-Extra: dev
 Provides-Extra: docs
 
-# diffaaable 1.0.1
+# diffaaable 1.1.1
 
 ![](docs/assets/diffaaable.png)
 
@@ -37,7 +39,7 @@ A detailed derivation of the used matrix expressions is provided in the appendix
 Under the hood `diffaaable` uses the AAA implementation of [`baryrat`](https://github.com/c-f-h/baryrat).
 Additionaly the following application specific extensions to the AAA algorithm are included:
 
-- **Adaptive**: Adaptive refinement strategy to minimize the number of function evaluation needed to precisely locate poles within some domain
+- **Adaptive**: Adaptive refinement strategy (called Iterative Sample Refinement (ISR) in the corresponding paper) to minimize the number of function evaluation needed to precisely locate poles within some domain
 - **Vectorial** (also referred to as set-valued): AAA algorithm acting on vector valued functions $\mathbf{f}(z)$ as presented in [^3].
 - **Lorentz**: Variant that enforces symmetric poles around the imaginary axis.
 - **Selective Refinement**: Use a divide and conquer theme to capture many pole simultaneously and accurately, by limiting the number of poles per AAA solve. Suggested in [^4].
@@ -58,7 +60,7 @@ When using this software package for scientific work please cite the associated
 +++
 
 [^1]: https://arxiv.org/pdf/2403.19404
-[^2]: Multiscat Resonances (to be publlished)
+[^2]: "A framework to compute resonances arising from multiple scattering", https://arxiv.org/abs/2409.05563
 [^3]: https://doi.org/10.1093/imanum/draa098
 [^4]: https://doi.org/10.48550/arXiv.2405.19582
 

{diffaaable-1.0.1 → diffaaable-1.1.1}/README.md

@@ -1,4 +1,4 @@
-# diffaaable 1.0.1
+# diffaaable 1.1.1
 
 ![](docs/assets/diffaaable.png)
 
@@ -7,7 +7,7 @@ A detailed derivation of the used matrix expressions is provided in the appendix
 Under the hood `diffaaable` uses the AAA implementation of [`baryrat`](https://github.com/c-f-h/baryrat).
 Additionaly the following application specific extensions to the AAA algorithm are included:
 
-- **Adaptive**: Adaptive refinement strategy to minimize the number of function evaluation needed to precisely locate poles within some domain
+- **Adaptive**: Adaptive refinement strategy (called Iterative Sample Refinement (ISR) in the corresponding paper) to minimize the number of function evaluation needed to precisely locate poles within some domain
 - **Vectorial** (also referred to as set-valued): AAA algorithm acting on vector valued functions $\mathbf{f}(z)$ as presented in [^3].
 - **Lorentz**: Variant that enforces symmetric poles around the imaginary axis.
 - **Selective Refinement**: Use a divide and conquer theme to capture many pole simultaneously and accurately, by limiting the number of poles per AAA solve. Suggested in [^4].
@@ -28,6 +28,6 @@ When using this software package for scientific work please cite the associated
 +++
 
 [^1]: https://arxiv.org/pdf/2403.19404
-[^2]: Multiscat Resonances (to be publlished)
+[^2]: "A framework to compute resonances arising from multiple scattering", https://arxiv.org/abs/2409.05563
 [^3]: https://doi.org/10.1093/imanum/draa098
 [^4]: https://doi.org/10.48550/arXiv.2405.19582

{diffaaable-1.0.1 → diffaaable-1.1.1}/diffaaable/__init__.py

@@ -1,6 +1,6 @@
 """diffaaable - JAX-differentiable AAA algorithm"""
 
-__version__ = "1.0.1"
+__version__ = "1.1.1"
 
 __all__ = ["aaa", "adaptive_aaa", "vectorial_aaa", "lorentz_aaa"]
 

{diffaaable-1.0.1 → diffaaable-1.1.1}/diffaaable/adaptive.py

@@ -113,6 +113,18 @@ def next_samples_heat(
 
 aaa = jax.tree_util.Partial(aaa)
 
+def mask(z_k, f_k, f_k_dot, cutoff):
+    m = np.abs(f_k)<cutoff    #filter out values, that have diverged too strongly
+    m = np.logical_and(m, ~np.isnan(f_k))   #filter out nans
+    m = np.logical_and(m, ~np.isnan(z_k))   #filter out nans
+
+    if m.ndim == 2:
+      m = np.all(m, axis=1)
+    z_k, f_k, f_k_dot = z_k[m], f_k[m], f_k_dot[m]
+
+    z_k, idx = np.unique(z_k, return_index=True) #filter out duplicates
+    return z_k, f_k[idx], f_k_dot[idx]
+
 def _adaptive_aaa(z_k_0: npt.NDArray,
                  f: callable,
                  evolutions: int = 2,
@@ -162,7 +174,7 @@ def _adaptive_aaa(z_k_0: npt.NDArray,
     f_k_dot = np.zeros_like(f_k)
 
   if cutoff is None:
-    cutoff = 1e10*np.median(np.abs(f_k))
+    cutoff = 1e10*np.nanmedian(np.abs(f_k))
 
   if domain is None:
     center = np.mean(z_k)
@@ -175,15 +187,9 @@ def _adaptive_aaa(z_k_0: npt.NDArray,
   if prev_z_n is None:
     prev_z_n = np.array([np.inf], dtype=complex)
 
-  def mask(z_k, f_k, f_k_dot):
-    m = np.abs(f_k)<cutoff
-    if m.ndim == 2:
-      m = np.all(m, axis=1) #filter out values, that have diverged too strongly
-    return z_k[m], f_k[m], f_k_dot[m]
-
   key = random.key(0)
   for i in range(evolutions):
-    z_k, f_k, f_k_dot = mask(z_k, f_k, f_k_dot)
+    z_k, f_k, f_k_dot = mask(z_k, f_k, f_k_dot, cutoff=cutoff)
     z_j, f_j, w_j, z_n = aaa(z_k, f_k, tol, mmax)
 
     if i==evolutions-1:
@@ -211,7 +217,7 @@ def _adaptive_aaa(z_k_0: npt.NDArray,
     f_k = np.concatenate([f_k, f_k_new])
     f_k_dot = np.concatenate([f_k_dot, f_k_dot_new])
 
-  z_k, f_k, f_k_dot = mask(z_k, f_k, f_k_dot)
+  z_k, f_k, f_k_dot = mask(z_k, f_k, f_k_dot, cutoff=cutoff)
 
   if collect_tangents:
     return z_k, f_k, f_k_dot

{diffaaable-1.0.1 → diffaaable-1.1.1}/diffaaable/selective.py

@@ -65,11 +65,6 @@ def subdomains(domain: Domain, divide_horizontal: bool, center: complex=None):
     return [(subs[1][0], subs[0][1]), (subs[2][0], subs[3][1])]
   return   [(subs[2][0], subs[1][1]), (subs[3][0], subs[0][1])]
 
-
-def cutoff_mask(z_k, f_k, f_k_dot, cutoff):
-  m = np.abs(f_k)<cutoff #filter out values, that have diverged too strongly
-  return z_k[m], f_k[m], f_k_dot[m]
-
 def plot_domain(domain: Domain, size: float=1):
   left_up =    domain[0].real + 1j*domain[1].imag
   right_down = domain[1].real + 1j*domain[0].imag
@@ -82,7 +77,6 @@ def plot_domain(domain: Domain, size: float=1):
 def all_poles_known(poles, prev, tol):
   if prev is None or len(prev)!=len(poles):
     return False
-  #return True
 
   dist = np.abs(poles[:, None] - prev[None, :])
   check = np.all(np.any(dist < tol, axis=1))
@@ -110,12 +104,9 @@ def selective_refinement_aaa(f: callable,
   TODO: allow access to samples slightly outside of domain
   """
 
-  print(f"start domain '{debug_name}', {Dmax=}")
+  # print(f"start domain '{debug_name}', {Dmax=}")
   folder = f"debug_out/{debug_name:0<33}"
   domain_size = np.abs(domain[1]-domain[0])/2
-  size = domain_size/2 # for plotting
-  #plot_rect = plot_domain(domain, size=30)
-  #color = plot_rect[0].get_color()
 
   if cutoff is None:
     cutoff = np.inf
@@ -135,11 +126,13 @@ def selective_refinement_aaa(f: callable,
       z_k = np.append(z_k, z_k_new)
 
       eval_count += len(z_k_new)
-      print(f"new eval: {eval_count}")
+      # print(f"new eval: {eval_count}")
     eval_count -= len(z_k)
     z_j, f_j, w_j, z_n, z_k, f_k = adaptive_aaa(
       z_k, f, f_k_0=f_k, evolutions=N*16, tol=tol_aaa,
-      domain=reduced_domain(domain, 1.07), radius=4*domain_size/(N), #NOTE: actually increased domain :/
+      domain=reduced_domain(domain, 1.07), radius=4*domain_size/(N),
+      # NOTE: reduced domain with a factor larger than 1
+      # actually increases domain size
       return_samples=True, sampling=sampling, cutoff=np.inf
     )
     # TODO pass down samples in buffer zone
@@ -156,18 +149,14 @@ def selective_refinement_aaa(f: callable,
     except onp.linalg.LinAlgError as e:
       z_n = z_j = f_j = w_j = np.empty((0,))
 
-  print(f"domain '{debug_name}' done: {domain} ->  eval: {eval_count}")
+  # print(f"domain '{debug_name}' done: {domain} ->  eval: {eval_count}")
   poles = z_n[domain_mask(domain, z_n)]
 
   if (Dmax == 0 or
     (len(poles)<=max_poles and all_poles_known(poles, suggestions, tol_pol))):
 
-    #plt.scatter(poles.real, poles.imag, color = color, marker="x")#, s=size*3, linewidths=size/2)
-    print("I am done here")
-
     res = residues(z_j, f_j, w_j, poles)
     return poles, res, eval_count
-  #plt.scatter(poles.real, poles.imag, color = color, marker="+", s=0.2, zorder=3)#, s=size, linewidths=size/6)
 
   subs = subdomains(domain, divide_horizontal)
 
@@ -190,8 +179,4 @@ def selective_refinement_aaa(f: callable,
     pol = np.append(pol, p)
     res = np.append(res, r)
     eval_count += e
-  # if len(pol) > 0:
-  #   plt.xlim(domain[0].real, domain[1].real)
-  #   plt.ylim(domain[0].imag, domain[1].imag)
-  #   plt.savefig(f"debug_out/{debug_name:0<33}.png")
   return pol, res, eval_count

{diffaaable-1.0.1 → diffaaable-1.1.1}/pyproject.toml

@@ -22,9 +22,7 @@ dependencies = [
   "jax",
   "jaxlib",
   "baryrat",
-  "jaxopt",
-  "pytest",
-  "pytest-benchmark"
+  "jaxopt"
 ]
 description = "JAX-differentiable AAA algorithm"
 keywords = ["python"]
@@ -32,14 +30,18 @@ license = {file = "LICENSE"}
 name = "diffaaable"
 readme = "README.md"
 requires-python = ">=3.9"
-version = "1.0.1"
+version = "1.1.1"
 
 [project.optional-dependencies]
 dev = [
+  "tbump>=6.11.0",
+  "towncrier",
   "pre-commit",
   "pytest",
   "pytest-cov",
-  "pytest_regressions"
+  "pytest_regressions",
+  "pytest-benchmark",
+  "matplotlib"
 ]
 docs = [
   "jupytext",
@@ -135,7 +137,7 @@ message_template = "Bump to {new_version}"
 tag_template = "v{new_version}"
 
 [tool.tbump.version]
-current = "1.0.1"
+current = "1.1.1"
 # Example of a semver regexp.
 # Make sure this matches current_version before
 # using tbump