LoopStructural 1.0.1__zip

Miniconda/envs/loop/Lib/site-packages/LoopStructural/interpolators/discrete_interpolator.py

@@ -0,0 +1,551 @@
+"""
+Discrete interpolator base for least squares
+"""
+import logging
+
+import numpy as np
+from scipy.sparse import coo_matrix, bmat, eye
+from scipy.sparse import linalg as sla
+
+from LoopStructural.interpolators.geological_interpolator import \
+    GeologicalInterpolator
+
+logger = logging.getLogger(__name__)
+
+
+class DiscreteInterpolator(GeologicalInterpolator):
+    """
+
+    """
+    def __init__(self, support):
+        """
+        Base class for a discrete interpolator e.g. piecewise linear or finite difference which is
+        any interpolator that solves the system using least squares approximation
+
+        Parameters
+        ----------
+        support 
+            A discrete mesh with, nodes, elements, etc
+        """
+        GeologicalInterpolator.__init__(self)
+        self.B = []
+        self.support = support
+        self.region_function = None
+        self.region = np.arange(0, support.n_nodes)
+        self.region_map = np.zeros(support.n_nodes).astype(int)
+        # self.region_map[self.region] = np.array(range(0,
+        # len(self.region_map[self.region])))
+        self.nx = len(self.support.nodes[self.region])
+        self.shape = 'rectangular'
+        if self.shape == 'square':
+            self.B = np.zeros(self.nx)
+        self.c_ = 0
+        self.A = []  # sparse matrix storage coo format
+        self.col = []
+        self.row = []  # sparse matrix storage
+        self.solver = None
+        self.eq_const_C = []
+        self.eq_const_row = []
+        self.eq_const_col = []
+        self.eq_const_d = []
+        self.eq_const_c_ = 0
+        self.constraints = {}
+        self.interpolation_weights= {}
+
+    def set_property_name(self, propertyname):
+        """
+        Set the property name attribute, this is usually used to
+        save the property on the support
+
+        Parameters
+        ----------
+        propertyname
+
+        Returns
+        -------
+
+        """
+        self.propertyname = propertyname
+
+    def set_region(self, region=None):
+        """
+        Set the region of the support the interpolator is working on
+
+        Parameters
+        ----------
+        region - function(position)
+            return true when in region, false when out
+
+        Returns
+        -------
+
+        """
+        # evaluate the region function on the support to determine
+        # which nodes are inside update region map and degrees of freedom
+        self.region_function = region
+        self.region = region(self.support.nodes)
+        self.region_map = np.zeros(self.support.n_nodes).astype(int)
+        self.region_map[self.region] = np.array(
+            range(0, len(self.region_map[self.region])))
+        self.nx = len(self.support.nodes[self.region])
+        
+    def set_interpolation_weights(self, weights):
+        """
+        Set the interpolation weights dictionary
+
+        Parameters
+        ----------
+        weights - dictionary
+            Entry of new weights to assign to self.interpolation_weights
+
+        Returns
+        -------
+
+        """
+        for key in weights:
+            self.up_to_date = False
+            self.interpolation_weights[key] = weights[key]
+
+    def reset(self):
+        """
+        Reset the interpolation constraints
+
+        """
+        logger.debug("Resetting interpolation constraints")
+        self.c_ = 0
+        self.A = []  # sparse matrix storage coo format
+        self.col = []
+        self.row = []  # sparse matrix storage
+        self.eq_const_C = []
+        self.eq_const_row = []
+        self.eq_const_col = []
+        self.eq_const_d = []
+        self.eq_const_c_ = 0
+        self.B = []
+        self.n_constraints = 0
+
+    def add_constraints_to_least_squares(self, A, B, idc, name='undefined'):
+        """
+        Adds constraints to the least squares system. Automatically works
+        out the row
+        index given the shape of the input arrays
+
+        Parameters
+        ----------
+        A : numpy array / list
+            RxC numpy array of constraints where C is number of columns,R rows
+        B : numpy array /list
+            B values array length R
+        idc : numpy array/list
+            RxC column index
+
+        Returns
+        -------
+        list of constraint ids
+
+        """
+        A = np.array(A)
+        B = np.array(B)
+        idc = np.array(idc)
+        if np.any(np.isnan(idc)) or np.any(np.isnan(A)) or np.any(np.isnan(B)):
+            logger.warning("Constraints contain nan not adding constraints")
+            return
+        nr = A.shape[0]
+        if len(A.shape) > 2:
+            nr = A.shape[0] * A.shape[1]
+        rows = np.arange(0, nr).astype(int)
+        rows += self.c_
+        constraint_ids = rows.copy()
+
+        if name in self.constraints:
+            self.constraints[name] = np.hstack([self.constraints[name],
+                                                constraint_ids])
+        if name not in self.constraints:
+            self.constraints[name] = constraint_ids
+        rows = np.tile(rows, (A.shape[-1], 1)).T
+
+        self.c_ += nr
+        if self.shape == 'rectangular':
+            # don't add operator where it is = 0 to the sparse matrix!
+            A = A.flatten()
+            rows = rows.flatten()
+            idc = idc.flatten()
+            B = B.flatten()
+            mask = A == 0
+            self.A.extend(A[~mask].tolist())
+            self.row.extend(rows[~mask].tolist())
+            self.col.extend(idc[~mask].tolist())
+            self.B.extend(B.tolist())
+
+    def remove_constraints_from_least_squares(self, name='undefined',
+                                              constraint_ids=None):
+        """
+        Remove constraints from the least squares system using the constraint ids
+        which corresponds to the rows in the interpolation matrix.
+
+        Parameters
+        ----------
+        constraint_ids : np.array(dtype=int)
+            id of constraints to remove
+
+        Returns
+        -------
+
+        """
+
+        if constraint_ids is None:
+            constraint_ids = self.constraints[name]
+        print("Removing {} {} constraints from least squares".format(len(constraint_ids), name))
+        A = np.array(self.A)
+        B = np.array(self.B)
+        col = np.array(self.col)
+        row = np.array(self.row)
+        mask = np.any((row[:,None] == constraint_ids[None,:]) == True,
+                      axis=1)
+        # np.any((numbers[:, None] == np.array([0, 10, 30])[None, :]) == True,
+        #        axis=1)
+        bmask = np.ones(B.shape,dtype=bool)
+        bmask[constraint_ids] = 0
+        self.A = A[~mask].tolist()
+        self.B = B[bmask]
+        self.col = col[~mask].tolist()
+        rowmax = np.max(row[mask])
+        rowrange = rowmax-np.min(row[mask])
+        # row[np.logical_and(~mask,row>rowmax)] -= rowrange
+        return row[~mask]
+
+    def add_equality_constraints(self, node_idx, values):
+        """
+        Adds hard constraints to the least squares system. For now this just
+        sets
+        the node values to be fixed using a lagrangian.
+
+        Parameters
+        ----------
+        node_idx : numpy array/list
+            int array of node indexes
+        values : numpy array/list
+            array of node values
+
+        Returns
+        -------
+
+        """
+        # map from mesh node index to region node index
+        gi = np.zeros(self.support.n_nodes)
+        gi[:] = -1
+        gi[self.region] = np.arange(0, self.nx)
+        idc = gi[node_idx]
+        outside = ~(idc == -1)
+
+        self.eq_const_C.extend(np.ones(idc[outside].shape[0]).tolist())
+        self.eq_const_col.extend(idc[outside].tolist())
+        self.eq_const_row.extend((np.arange(0, idc[outside].shape[0])))
+        self.eq_const_d.extend(values[outside].tolist())
+        self.eq_const_c_ += idc[outside].shape[0]
+
+    def add_tangent_ctr_pts(self, w=1.0):
+        """
+
+        Parameters
+        ----------
+        w : double
+
+
+        Returns
+        -------
+
+        """
+        points = self.get_tangent_constraints()
+        if points.shape[0] > 1:
+            self.add_gradient_orthogonal_constraint(points[:,:3],points[:,3:6],w)
+
+    def build_matrix(self, square=True, damp=True):
+        """
+        Assemble constraints into interpolation matrix. Adds equaltiy
+        constraints
+        using lagrange modifiers if necessary
+
+        Parameters
+        ----------
+        damp: bool
+            Flag whether damping should be added to the diagonal of the matrix
+        Returns
+        -------
+        Interpolation matrix and B
+        """
+
+        logger.info("Interpolation matrix is %i x %i"%(self.c_,self.nx))
+        cols = np.array(self.col)
+        A = coo_matrix((np.array(self.A), (np.array(self.row), \
+                                           cols)), shape=(self.c_, self.nx),
+                       dtype=float)  # .tocsr()
+        B = np.array(self.B)
+        if not square:
+            logger.info("Using rectangular matrix, equality constraints are not used")
+            return A, B
+        AAT = A.T.dot(A)
+        BT = A.T.dot(B)
+        # add a small number to the matrix diagonal to smooth the results
+        # can help speed up solving, but might also introduce some errors
+
+        if self.eq_const_c_ > 0:
+            logger.info("Equality block is %i x %i"%(self.eq_const_c_,self.nx))
+            # solving constrained least squares using
+            # | ATA CT | |c| = b
+            # | C   0  | |y|   d
+            # where A is the interpoaltion matrix
+            # C is the equality constraint matrix
+            # b is the interpolation constraints to be honoured
+            # in a least squares sense
+            # and d are the equality constraints
+            # c are the node values and y are the
+            # lagrange multipliers#
+            C = coo_matrix(
+                (np.array(self.eq_const_C), (np.array(self.eq_const_row),
+                                             np.array(self.eq_const_col))),
+                shape=(self.eq_const_c_, self.nx))
+            d = np.array(self.eq_const_d)
+            AAT = bmat([[AAT, C.T], [C, None]])
+            BT = np.hstack([BT, d])
+        if damp:
+            logger.info("Adding eps to matrix diagonal")
+            AAT += eye(AAT.shape[0]) * np.finfo('float').eps
+        return AAT, BT
+
+    def _solve_lu(self, A, B):
+        """
+        Call scipy LU decomoposition
+
+        Parameters
+        ----------
+        A : scipy square sparse matrix
+        B : numpy vector
+
+        Returns
+        -------
+
+        """
+        lu = sla.splu(A.tocsc())
+        sol = lu.solve(B)
+        return sol[:self.nx]
+
+    def _solve_lsqr(self, A, B, **kwargs):
+        """
+        Call scipy lsqr
+
+        Parameters
+        ----------
+        A : rectangular sparse matrix
+        B : vector
+
+        Returns
+        -------
+
+        """
+
+        lsqrargs = {}
+        # lsqrargs['tol'] = 1e-12
+        if 'iter_lim' in kwargs:
+            logger.info("Using %i maximum iterations" % kwargs['iter_lim'])
+            lsqrargs['iter_lim'] = kwargs['iter_lim']
+        if 'damp' in kwargs:
+            logger.info("Using damping coefficient")
+            lsqrargs['damp'] = kwargs['damp']
+        if 'atol' in kwargs:
+            logger.info('Using a tolerance of %f' % kwargs['atol'])
+            lsqrargs['atol'] = kwargs['atol']
+        if 'btol' in kwargs:
+            logger.info('Using btol of %f' % kwargs['btol'])
+            lsqrargs['btol'] = kwargs['btol']
+        if 'show' in kwargs:
+            lsqrargs['show'] = kwargs['show']
+        if 'conlim' in kwargs:
+            lsqrargs['conlim'] = kwargs['conlim']
+        return sla.lsqr(A,B, **lsqrargs)[0]
+
+    def _solve_chol(self, A, B):
+        """
+        Call suitesparse cholmod through scikitsparse
+        LINUX ONLY!
+
+        Parameters
+        ----------
+        A : scipy.sparse.matrix
+            square sparse matrix
+        B : numpy array
+            RHS of equation
+
+        Returns
+        -------
+
+        """
+        try:
+            from sksparse.cholmod import cholesky
+            factor = cholesky(A.tocsc())
+            return factor(B)[:self.nx]
+        except ImportError:
+            logger.warning("Scikit Sparse not installed try using cg instead")
+            return False
+
+    def _solve_cg(self, A, B, precon=None, **kwargs):
+        """
+        Call scipy conjugate gradient
+
+        Parameters
+        ----------
+        A : scipy.sparse.matrix
+            square sparse matrix
+        B : numpy vector
+        precon : scipy.sparse.matrix
+            a preconditioner for the conjugate gradient system
+        kwargs
+            kwargs to pass to scipy solve e.g. atol, btol, callback etc
+
+        Returns
+        -------
+        numpy array
+        """
+        cgargs = {}
+        cgargs['tol'] = 1e-12
+        cgargs['atol'] = 0
+        if 'maxiter' in kwargs:
+            logger.info("Using %i maximum iterations"%kwargs['maxiter'])
+            cgargs['maxiter'] = kwargs['maxiter']
+        if 'x0' in kwargs:
+            logger.info("Using starting guess")
+            cgargs['x0'] = kwargs['x0']
+        if 'tol' in kwargs:
+            logger.info('Using tolerance of %f'%kwargs['tol'])
+            cgargs['tol'] = kwargs['tol']
+        if 'atol' in kwargs:
+            logger.info('Using atol of %f'%kwargs['atol'])
+            cgargs['atol'] = kwargs['atol']
+        if 'callback' in kwargs:
+            cgargs['callback'] = kwargs['callback']
+        if precon is not None:
+            cgargs['M'] = precon(A)
+        return sla.cg(A, B, **cgargs)[0][:self.nx]
+
+    def _solve_pyamg(self, A, B):
+        """
+        Solve least squares system using pyamg algorithmic multigrid solver
+
+        Parameters
+        ----------
+        A :  scipy.sparse.matrix
+        B : numpy array
+
+        Returns
+        -------
+
+        """
+        import pyamg
+        return pyamg.solve(A, B, verb=False)[:self.nx]
+
+    def _solve(self, solver='cg', **kwargs):
+        """
+        Main entry point to run the solver and update the node value
+        attribute for the
+        discreteinterpolator class
+
+        Parameters
+        ----------
+        solver : string
+            solver e.g. cg, lu, chol, custom
+        kwargs
+            kwargs for solver e.g. maxiter, preconditioner etc, damping for
+        
+        Returns
+        -------
+        bool
+            True if the interpolation is run
+
+        """
+        self.c = np.zeros(self.support.n_nodes)
+        self.c[:] = np.nan
+        damp = True
+        if 'damp' in kwargs:
+            damp = kwargs['damp']
+        if solver == 'lu':
+            logger.info("Forcing matrix damping for LU")
+            damp = True
+        if solver == 'lsqr':
+            A, B =  self.build_matrix(False)
+        else:
+            A, B = self.build_matrix(damp=damp)
+
+        # run the chosen solver
+        if solver == 'cg':
+            logger.info("Solving using conjugate gradient")
+            self.c[self.region] = self._solve_cg(A, B, **kwargs)
+        if solver == 'chol':
+            self.c[self.region] = self._solve_chol(A, B)
+        if solver == 'lu':
+            logger.info("Solving using scipy LU")
+            self.c[self.region] = self._solve_lu(A, B)
+        if solver == 'pyamg':
+            try:
+                logger.info("Solving with pyamg solve")
+                self.c[self.region] = self._solve_pyamg(A, B)
+            except ImportError:
+                logger.warn("Pyamg not installed using cg instead")
+                self.c[self.region] = self._solve_cg(A, B)
+        if solver == 'lsqr':
+            self.c[self.region] = self._solve_lsqr(A, B, **kwargs)
+        if solver == 'external':
+            logger.warning("Using external solver")
+            self.c[self.region] = kwargs['external'](A, B)[:self.nx]
+        # check solution is not nan
+        self.support.properties[self.propertyname] = self.c
+        if np.all(self.c == np.nan):
+            logger.warning("Solver not run, no scalar field")
+        # if solution is all 0, probably didn't work
+        if np.all(self.c[self.region] == 0):
+            logger.warning("No solution, scalar field 0. Add more data.")
+
+    def update(self):
+        """
+        Check if the solver is up to date, if not rerun interpolation using
+        the previously used solver. If the interpolation has not been run
+        before it will
+        return False
+
+        Returns
+        -------
+        bool
+
+        """
+        if self.solver is None:
+            logging.debug("Cannot rerun interpolator")
+            return False
+        if not self.up_to_date:
+            self.setup_interpolator()
+            return self._solve(self.solver)
+
+    def evaluate_value(self, evaluation_points):
+        evaluation_points = np.array(evaluation_points)
+        evaluated = np.zeros(evaluation_points.shape[0])
+        mask = np.any(evaluation_points == np.nan, axis=1)
+
+        if evaluation_points[~mask, :].shape[0] > 0:
+            evaluated[~mask] = self.support.evaluate_value(
+                evaluation_points[~mask], self.propertyname)
+        return evaluated
+
+    def evaluate_gradient(self, evaluation_points):
+        """
+        Evaluate the gradient of the scalar field at the evaluation points
+        Parameters
+        ----------
+        evaluation_points : np.array
+            xyz locations to evaluate the gradient
+
+        Returns
+        -------
+
+        """
+        if evaluation_points.shape[0] > 0:
+            return self.support.evaluate_gradient(evaluation_points,
+                                                  self.propertyname)
+        return np.zeros((0, 3))