Merge branch 'master' into cln_process_multiple

Author: Jacob-Stevens-Haas

pysindy/differentiation/sindy_derivative.py

@@ -51,7 +51,7 @@ def set_params(self, **params):
             # Simple optimization to gain speed (inspect is slow)
             return self
         else:
-            self.kwargs.update(params)
+            self.kwargs.update(params["kwargs"])
 
         return self
 

pysindy/feature_library/base.py

@@ -13,13 +13,12 @@
 from sklearn.utils.validation import check_is_fitted
 
 from ..utils import AxesArray
-from ..utils import DefaultShapedInputsMixin
+from ..utils import comprehend_axes
 from ..utils import validate_no_reshape
 from ..utils import wrap_axes
-from ..utils.axes import ax_time_to_ax_sample
 
 
-class BaseFeatureLibrary(DefaultShapedInputsMixin, TransformerMixin):
+class BaseFeatureLibrary(TransformerMixin):
     """
     Base class for feature libraries.
 
@@ -187,17 +186,12 @@ def func(self, x, *args, **kwargs):
             return wrapped_func(self, x, *args, **kwargs)
         else:
             if not sparse.issparse(x):
-                x = AxesArray(x, self.comprehend_axes(x))
-                x = ax_time_to_ax_sample(x)
+                x = AxesArray(x, comprehend_axes(x))
                 reconstructor = np.array
             else:  # sparse arrays
                 reconstructor = type(x)
-                axes = self.comprehend_axes(x)
-                wrap_axes(axes)(x)
-                # Can't use x = ax_time_to_ax_sample(x) b/c that creates
-                # an AxesArray
-                x.ax_sample = x.ax_time
-                x.ax_time = None
+                axes = comprehend_axes(x)
+                wrap_axes(axes, x)
             result = wrapped_func(self, [x], *args, **kwargs)
             if isinstance(result, Sequence):  # e.g. transform() returns x
                 return reconstructor(result[0])
@@ -332,8 +326,8 @@ def transform(self, x_full):
                 xp[..., start_feature_index:end_feature_index] = lib.transform([x])[0]
 
                 current_feat += lib_n_output_features
-
-            xp_full = xp_full + [AxesArray(xp, self.comprehend_axes(xp))]
+            xp = AxesArray(xp, comprehend_axes(xp))
+            xp_full.append(xp)
         if self.library_ensemble:
             xp_full = self._ensemble(xp_full)
         return xp_full
@@ -565,7 +559,8 @@ def transform(self, x_full):
 
                     current_feat += lib_i_n_output_features * lib_j_n_output_features
 
-            xp_full = xp_full + [AxesArray(xp, self.comprehend_axes(xp))]
+            xp = AxesArray(xp, comprehend_axes(xp))
+            xp_full.append(xp)
         if self.library_ensemble:
             xp_full = self._ensemble(xp_full)
         return xp_full

pysindy/feature_library/custom_library.py

@@ -8,6 +8,7 @@
 from sklearn.utils.validation import check_is_fitted
 
 from ..utils import AxesArray
+from ..utils import comprehend_axes
 from .base import BaseFeatureLibrary
 from .base import x_sequence_or_item
 
@@ -229,7 +230,8 @@ def transform(self, x_full):
                     xp[..., library_idx] = f(*[x[..., j] for j in c])
                     library_idx += 1
 
-            xp_full = xp_full + [AxesArray(xp, self.comprehend_axes(xp))]
+            xp = AxesArray(xp, comprehend_axes(xp))
+            xp_full.append(xp)
         if self.library_ensemble:
             xp_full = self._ensemble(xp_full)
         return xp_full

pysindy/feature_library/fourier_library.py

@@ -3,6 +3,7 @@
 from sklearn.utils.validation import check_is_fitted
 
 from ..utils import AxesArray
+from ..utils import comprehend_axes
 from .base import BaseFeatureLibrary
 from .base import x_sequence_or_item
 
@@ -173,8 +174,8 @@ def transform(self, x_full):
                     if self.include_cos:
                         xp[..., idx] = np.cos((i + 1) * x[..., j])
                         idx += 1
-
-            xp_full = xp_full + [AxesArray(xp, self.comprehend_axes(xp))]
+            xp = AxesArray(xp, comprehend_axes(xp))
+            xp_full.append(xp)
         if self.library_ensemble:
             xp_full = self._ensemble(xp_full)
         return xp_full

pysindy/feature_library/generalized_library.py

@@ -1,5 +1,3 @@
-from typing import Type
-
 import numpy as np
 from sklearn import __version__
 from sklearn.utils.validation import check_is_fitted
@@ -130,12 +128,10 @@ def __init__(
             if weak_libraries:
                 self.validate_input = libraries[weak_libraries].validate_input
                 self.calc_trajectory = libraries[weak_libraries].calc_trajectory
-                self.comprehend_axes = libraries[weak_libraries].comprehend_axes
                 self.spatiotemporal_grid = libraries[weak_libraries].spatiotemporal_grid
             elif pde_libraries:
                 self.validate_input = libraries[pde_libraries].validate_input
                 self.calc_trajectory = libraries[pde_libraries].calc_trajectory
-                self.comprehend_axes = libraries[pde_libraries].comprehend_axes
                 self.spatial_grid = libraries[pde_libraries].spatial_grid
         else:
             raise ValueError(
@@ -246,23 +242,6 @@ def fit(self, x_full, y=None):
 
         return self
 
-    def has_type(self, libtype: Type, exclusively=False) -> bool:
-        """Checks whether this library has a specific library type.
-
-        Parameters
-        ----------
-        libtype : A type of feature library
-        exclusively: whether to check all libraries
-
-        Returns
-        -------
-        Bool indicating whether specific library type is present
-        """
-        has_inst = map(lambda lib: isinstance(lib, libtype), self.libraries_)
-        if exclusively:
-            return all(has_inst)
-        return any(has_inst)
-
     @x_sequence_or_item
     def transform(self, x_full):
         """Transform data with libs provided below.
@@ -283,13 +262,9 @@ def transform(self, x_full):
 
         xp_full = []
         for x in x_full:
-            # n_samples = x.shape[x.ax_sample]
             n_features = x.shape[x.ax_coord]
             shape = np.array(x.shape)
 
-            # if isinstance(self.libraries_[0], WeakPDELibrary):
-            #     n_samples = self.libraries_[0].K * self.libraries_[0].num_trajectories
-
             if float(__version__[:3]) >= 1.0:
                 n_input_features = self.n_features_in_
             else:

pysindy/feature_library/pde_library.py

@@ -7,6 +7,7 @@
 from sklearn.utils.validation import check_is_fitted
 
 from ..utils import AxesArray
+from ..utils import comprehend_axes
 from .base import BaseFeatureLibrary
 from .base import x_sequence_or_item
 from pysindy.differentiation import FiniteDifference
@@ -197,10 +198,6 @@ def __init__(
             spatiotemporal_grid = np.reshape(
                 spatiotemporal_grid, (len(spatiotemporal_grid), 1)
             )
-        self.grid_dims = spatiotemporal_grid.shape[:-1]
-        self.spatial_grid_dims = self.spatial_grid.shape[:-1]
-        self.grid_ndim = len(spatiotemporal_grid.shape[:-1])
-        self.spatial_grid_ndim = len(self.spatial_grid_dims)
 
         # if want to include temporal terms -> range(len(dims))
         if self.implicit_terms:
@@ -222,7 +219,9 @@ def __init__(
 
         self.num_derivatives = num_derivatives
         self.multiindices = multiindices
-        self.spatiotemporal_grid = spatiotemporal_grid
+        self.spatiotemporal_grid = AxesArray(
+            spatiotemporal_grid, comprehend_axes(spatiotemporal_grid)
+        )
 
     @staticmethod
     def _combinations(n_features, n_args, interaction_only):
@@ -276,7 +275,13 @@ def get_feature_names(self, input_features=None):
         def derivative_string(multiindex):
             ret = ""
             for axis in range(self.ind_range):
-                if (axis == self.grid_ndim - 1) and self.implicit_terms:
+                if self.implicit_terms and (
+                    axis
+                    in [
+                        self.spatiotemporal_grid.ax_time,
+                        self.spatiotemporal_grid.ax_sample,
+                    ]
+                ):
                     str_deriv = "t"
                 else:
                     str_deriv = str(axis + 1)
@@ -392,11 +397,11 @@ def transform(self, x_full):
             library_derivatives = np.empty(shape, dtype=x.dtype)
             library_idx = 0
             for multiindex in self.multiindices:
-                derivs = x.copy()
+                derivs = x
                 for axis in range(self.ind_range):
                     if multiindex[axis] > 0:
                         s = [0 for dim in self.spatiotemporal_grid.shape]
-                        s[axis] = slice(self.grid_dims[axis])
+                        s[axis] = slice(self.spatiotemporal_grid.shape[axis])
                         s[-1] = axis
 
                         derivs = FiniteDifference(
@@ -459,7 +464,8 @@ def transform(self, x_full):
                     shape,
                 )
                 library_idx += n_library_terms * self.num_derivatives * n_features
-            xp_full = xp_full + [AxesArray(xp, self.comprehend_axes(xp))]
+            xp = AxesArray(xp, comprehend_axes(xp))
+            xp_full.append(xp)
         if self.library_ensemble:
             xp_full = self._ensemble(xp_full)
         return xp_full

pysindy/feature_library/polynomial_library.py

@@ -10,6 +10,7 @@
 from sklearn.utils.validation import check_is_fitted
 
 from ..utils import AxesArray
+from ..utils import comprehend_axes
 from ..utils import wrap_axes
 from .base import BaseFeatureLibrary
 from .base import x_sequence_or_item
@@ -229,15 +230,11 @@ def transform(self, x_full):
         for x in x_full:
             if sparse.issparse(x) and x.format not in ["csr", "csc"]:
                 # create new with correct sparse
-                axes = self.comprehend_axes(x)
+                axes = comprehend_axes(x)
                 x = x.asformat("csr")
-                wrap_axes(axes)(x)
-                # Can't use x = ax_time_to_ax_sample(x) b/c that creates
-                # an AxesArray
-                x.ax_sample = x.ax_time
-                x.ax_time = None
+                wrap_axes(axes, x)
 
-            n_samples = x.shape[x.ax_sample]
+            n_samples = x.shape[x.ax_time]
             n_features = x.shape[x.ax_coord]
             if float(__version__[:3]) >= 1.0:
                 if n_features != self.n_features_in_:

pysindy/feature_library/weak_pde_library.py

@@ -831,28 +831,6 @@ def transform(self, x_full):
         """
         check_is_fitted(self)
 
-        # x = check_array(x)
-        #
-        # n_samples_original_full, n_features = x.shape
-        # n_samples_original = n_samples_original_full // self.num_trajectories
-        #
-        # if float(__version__[:3]) >= 1.0:
-        #     if n_features != self.n_features_in_:
-        #         raise ValueError("x shape does not match training shape")
-        # else:
-        #     if n_features != self.n_input_features_:
-        #         raise ValueError("x shape does not match training shape")
-        #
-        # if self.spatiotemporal_grid is not None:
-        #     n_samples = self.K
-        #     n_samples_full = self.K * self.num_trajectories
-        #
-        # xp_full = np.empty(
-        #     (self.num_trajectories, n_samples, self.n_output_features_), dtype=x.dtype
-        # )
-        # x_full = np.reshape(
-        #     x, np.concatenate([[self.num_trajectories], self.grid_dims, [n_features]])
-        # )
         xp_full = []
         for x in x_full:
             n_features = x.shape[x.ax_coord]
@@ -871,12 +849,6 @@ def transform(self, x_full):
             library_functions = np.empty((self.K, n_library_terms), dtype=x.dtype)
 
             # Evaluate the functions on the indices of domain cells
-            # x_shaped = np.reshape(
-            #     x,
-            #     np.concatenate([self.spatiotemporal_grid.shape[:-1], [x.shape[-1]]]),
-            # )
-            # dims = np.array(x_shaped.shape)
-            # dims[-1] = n_library_terms
             funcs = np.zeros((*x.shape[:-1], n_library_terms))
             func_idx = 0
             for f in self.functions:

pysindy/optimizers/base.py

@@ -363,6 +363,5 @@ def _drop_random_samples(
     rand_inds = choice(range(n_samples), n_subset, replace=replace)
     x_new = np.take(x, rand_inds, axis=x.ax_sample)
     x_dot_new = np.take(x_dot, rand_inds, axis=x.ax_sample)
-    # x_dot_new = np.take(x_dot, rand_inds, axis=x_dot.ax_sample)
 
     return x_new, x_dot_new