Tidy up polynomials

Author: alexgdebeer

deep_tensor/polynomials/cdf_1d.py

@@ -34,12 +34,8 @@ def nodes(self, value: Tensor) -> None:
         return
 
     @property 
-    @abc.abstractmethod 
-    def cardinality(self) -> Tensor:
-        """The number of nodes associated with the polynomial basis of
-        the CDF.
-        """
-        pass
+    def cardinality(self) -> int:
+        return self.nodes.numel()
 
     @property
     @abc.abstractmethod

deep_tensor/polynomials/piecewise/lagrange_1.py

@@ -51,15 +51,6 @@ def __init__(self, num_elems: int):
         self.mass_R = torch.linalg.cholesky(mass).T
         return
 
-    # @property
-    # def nodes(self) -> Tensor:
-    #     return self._nodes
-    
-    # @nodes.setter 
-    # def nodes(self, value: Tensor) -> None:
-    #     self._nodes = value 
-    #     return
-    
     @property 
     def mass_R(self) -> Tensor:
         return self._mass_R

deep_tensor/polynomials/piecewise/lagrange_p.py

@@ -185,11 +185,11 @@ def int_W(self, value: Tensor) -> None:
         return
 
     @property
-    def cardinality(self):
+    def cardinality(self) -> int:
         return self.nodes.numel()
 
     @property
-    def domain(self):
+    def domain(self) -> Tensor:
         return torch.tensor([-1.0, 1.0])
 
     @property 

deep_tensor/polynomials/piecewise/piecewise.py

@@ -82,7 +82,7 @@ def eval_measure(self, ls: Tensor) -> Tensor:
     def eval_log_measure(self, ls: Tensor) -> Tensor:
         return torch.full(ls.shape, -self.domain_size.log())
 
-    def eval_measure_deriv(obj, ls: Tensor) -> Tensor:
+    def eval_measure_deriv(self, ls: Tensor) -> Tensor:
         return torch.zeros_like(ls)
 
     def eval_log_measure_deriv(self, ls: Tensor) -> Tensor:

deep_tensor/polynomials/spectral/chebyshev_1st.py

@@ -1,3 +1,5 @@
+import math
+
 import torch
 from torch import Tensor
 
@@ -91,7 +93,7 @@ def eval_log_measure(self, ls: Tensor) -> Tensor:
         self._check_in_domain(ls)
         ts = 1.0 - ls.square()
         ts[ts < EPS] = EPS
-        return -0.5 * ts.log() - torch.tensor(torch.pi).log()
+        return -0.5 * torch.log(ts) - math.log(torch.pi)
 
     def eval_log_measure_deriv(self, ls: Tensor) -> Tensor:
         self._check_in_domain(ls)

deep_tensor/polynomials/spectral/chebyshev_1st_trigo_cdf.py

@@ -1,3 +1,5 @@
+from typing import Tuple
+
 import torch 
 from torch import Tensor 
 
@@ -59,7 +61,7 @@ def eval_int_basis(self, thetas: Tensor) -> Tensor:
         ))
         return int_pws
 
-    def eval_int_basis_newton(self, thetas: Tensor) -> Tensor:
+    def eval_int_basis_newton(self, thetas: Tensor) -> Tuple[Tensor, Tensor]:
         int_pws = self.eval_int_basis(thetas)
         thetas = thetas[:, None]
         derivs = self.norm * torch.cos(thetas * self.n) / torch.pi

deep_tensor/polynomials/spectral/chebyshev_2nd.py

@@ -1,3 +1,5 @@
+import math
+
 import torch 
 from torch import Tensor
 from torch.distributions.beta import Beta
@@ -87,7 +89,7 @@ def eval_measure(self, ls: Tensor) -> Tensor:
     def eval_log_measure(self, ls: Tensor) -> Tensor:
         self._check_in_domain(ls)
         ts = 1.0 - ls.square()
-        logws = 0.5 * ts.log() + torch.tensor(2.0/torch.pi).log()
+        logws = 0.5 * ts.log() + math.log(2.0/torch.pi)
         return logws
 
     def eval_measure_deriv(self, ls: Tensor) -> Tensor:

deep_tensor/polynomials/spectral/chebyshev_2nd_trigo_cdf.py

@@ -1,3 +1,5 @@
+from typing import Tuple
+
 import torch 
 from torch import Tensor 
 
@@ -60,7 +62,7 @@ def eval_int_basis(self, thetas: Tensor) -> Tensor:
 
         return ps
 
-    def eval_int_basis_newton(self, thetas: Tensor) -> Tensor:
+    def eval_int_basis_newton(self, thetas: Tensor) -> Tuple[Tensor, Tensor]:
 
         thetas = thetas[:, None]
 

deep_tensor/polynomials/spectral/jacobi_11.py

@@ -1,3 +1,5 @@
+import math
+
 import torch 
 from torch import Tensor 
 from torch.distributions.beta import Beta
@@ -45,7 +47,7 @@ def eval_measure(self, ls: Tensor) -> Tensor:
         return ws
 
     def eval_log_measure(self, ls: Tensor) -> Tensor:
-        ws = (1.0 - ls.square()).log() + torch.tensor(0.75).log()
+        ws = (1.0 - ls.square()).log() + math.log(0.75)
         return ws
 
     def eval_measure_deriv(self, ls: Tensor) -> Tensor:

deep_tensor/polynomials/spectral/recurr.py

@@ -1,5 +1,5 @@
 import abc
-from typing import Tuple
+from typing import Tuple, Union
 
 import torch
 from torch import Tensor
@@ -15,7 +15,7 @@ def __init__(
         a: Tensor,
         b: Tensor,
         c: Tensor,
-        norm: float | Tensor
+        norm: Tensor
     ):
         """Class for spectral polynomials for which the three-term 
         recurrence relation is known. This relation takes the form