fix: nonlinear will replace discrete class

Will use a similar parent class for the child classes but adding
non linear constraints

LoopStructural/interpolators/nonlinear_interpolator.py

@@ -0,0 +1,58 @@
+# from LoopStructural.interpolators import DiscreteInterpolator
+
+# class NonLinearInterpolator(DiscreteInterpolator):
+#     def __init__(self, support):
+#         """
+#         """
+#         super().__init__(self, support)
+#         self._non_linear_constraints = {}
+
+#     def add_non_linear_constraint(self,):
+#         self._non_linear_constraints[name] = constraint
+
+#     def _solve(self,solver,niter=10,**kwargs):
+#         A_base, B_base = feature.interpolator.build_matrix()
+#         for i in range(niter):
+#             for constraint in self._non_linear_constraints:
+#                 ATA, ATB = constraint()
+#                 A = ATA+A_base
+#                 B = ATB+B_base
+#         def iterative_norm(feature,niter,w=0.1):
+#     values = np.zeros((niter,100,100))
+#     norms = np.zeros((niter,100,100))
+#     pts = np.array(np.meshgrid(np.linspace(model.bounding_box[0,0],model.bounding_box[1,0],100),np.linspace(model.bounding_box[0,1],model.bounding_box[1,1],100))).T.reshape((-1,2))
+#     pts = np.hstack([pts,np.zeros((pts.shape[0],1))])
+#     feature.interpolator.solve_system(solver='cg')
+
+#     v = feature.evaluate_value(pts)
+#     values[0,:,:] = v.reshape((100,100))
+
+#     norm = np.linalg.norm(feature.evaluate_gradient(pts),axis=1)
+#     norms[0,:,:] = norm.reshape((100,100))
+#     print('Average gradient norm: {} std: {}'.format(np.nanmean(norm),np.nanstd(norm)))
+#     A_base, B_base = feature.interpolator.build_matrix()
+
+#     for i in range(1,niter):
+#         print('Adding constant norm: {} / {}'.format(i+1,niter))
+#         ATA, ATB = constant_norm_constraint(feature)
+
+#         ATA*=w*((i**2+1)/(niter**2))
+
+
+#         A = ATA+A_base
+#         B = ATB+B_base
+#         from scipy.sparse.linalg import cg, splu
+
+#         soln = cg(A,B,tol=feature.builder.build_arguments['tol'])
+#         feature.interpolator.c = soln[0]
+#         v = feature.evaluate_value(pts)
+
+#         values[i,:,:] = v.reshape((100,100))
+
+#         norm = np.linalg.norm(feature.evaluate_gradient(pts),axis=1)
+#         norms[i,:,:] = norm.reshape((100,100))
+
+#         print('Average gradient norm: {} std: {}'.format(np.nanmean(norm),np.nanstd(norm)))
+
+#     return values, norms
+