WIP

geometricalgebra/cga3d.py

@@ -178,3 +178,35 @@ def get_origin_pose(weight: float = 1.0):
     return Vector.concatenate(
         [e_0[None], weight * TETRAHEDRON / 2, weight * Vector.from_scalar(np.sqrt(3), pseudo=True)[None]]
     )
+
+
+def classify(vector: Vector, eps=1e-8):
+    """Classify the vector
+
+    Args:
+        vector: the vector to classify
+
+    Returns:
+        the classification of the vector
+    """
+    if len(vector.grades) == 1:
+        if vector.grade == 0:
+            return "scalar"
+        if vector.grade == 1:
+            pass
+
+        if vector.grade == 5:
+            return "pseudo-scalar"
+
+
+        if vector.grades == {1}:
+            if vector.square_norm() < eps:
+                return "flat point"
+            else:
+                raise NotImplementedError()
+
+        if vector.is_flat():
+            pass
+
+    raise NotImplementedError("multi-grade vector")
+

geometricalgebra/utils.py

@@ -1,9 +1,14 @@
 """Collection of utilities"""
 
 from geometricalgebra.vector import FRAMEWORK
-
+from typing import Tuple
 
 def solve_gauss_normal_equations(a, b):
     xnp = FRAMEWORK.numpy
     at_a_inv = xnp.linalg.pinv(xnp.einsum("...jk,...ji->...ki", a, a))
     return xnp.einsum("...ij,...kj,...k->...i", at_a_inv, a, b)
+
+
+def reshape_last_dimensions(array, num_of_dim: int, new_shape: Tuple[int, ...]):
+    """Reshape a tensor with shape (..., 4, 3) to shape (..., 12)"""
+    return array.reshape([*array.shape[:-num_of_dim], *new_shape])

geometricalgebra/vector.py

@@ -19,7 +19,7 @@ class Framework(NamedTuple):
     softmax: Any
 
 
-ga_numpy = os.environ.get("GEOMETRICALGEBRA_NUMPY", "numpy")
+ga_numpy = os.environ.get("GEOMETRICALGEBRA_NUMPY", "tensorflow")
 if ga_numpy == "jax":
     import jax
     import jax.numpy as jnp
@@ -117,9 +117,6 @@ def algebra(cls):
     def _algebra(self):
         return self.algebra
 
-    # def framework(self):
-    #     return jnp
-
     def numpy(self):
         return self