geomfum.numerics package#

Submodules#

geomfum.numerics.eig module#

Module for eigenvalue solver.

class geomfum.numerics.eig.ScipyEigsh(spectrum_size=6, sigma=None, which='LM')[source]#

Bases: object

Sparse eigenvalue solver using SciPy’s ARPACK wrapper.

Parameters:

  • spectrum_size (int, optional) – Number of eigenvalues and eigenvectors to compute (default: 6).

  • sigma (float, optional) – Shift for shift-invert mode. If None, standard mode is used.

  • which (str, optional) – Which eigenvalues to find: ‘LM’ (largest magnitude), ‘SM’ (smallest magnitude), ‘LA’ (largest algebraic), ‘SA’ (smallest algebraic), etc. (default: ‘LM’).

geomfum.numerics.graph module#

Routines for working with graphs.

geomfum.numerics.graph.single_source_partial_dijkstra_path_length(graph, source, k, weight='weight')[source]#

Compute shortest-path distances from a source node to the k closest nodes.

Based on cumulative path cost, using an early-stopped Dijkstra’s algorithm.

The search terminates once k nodes (including the source itself) have been reached.

Parameters:

  • graph (networkx.Graph) – The input graph. Can be directed or undirected. Edge weights must be non-negative.

  • source (node) – The starting node for paths.

  • k (int) – Number of nodes to find distances to (including the source itself).

Returns:

length (dict) – Dict keyed by node to shortest path length from source.

geomfum.numerics.optimization module#

Optimization routines.

class geomfum.numerics.optimization.ScipyMinimize(method='L-BFGS-B', bounds=None, constraints=(), tol=None, callback=None, options=None, save_result=False)[source]#

Bases: object

Backend-agnostic wrapper for SciPy’s optimization routines.

Parameters:

  • method (str, optional) – Optimization algorithm (default: ‘L-BFGS-B’).

  • bounds (sequence, optional) – Bounds on variables for constrained methods.

  • constraints (dict or sequence of dict, optional) – Constraints definition.

  • tol (float, optional) – Tolerance for termination.

  • callback (callable, optional) – Function called after each iteration.

  • options (dict, optional) – Solver-specific options.

  • save_result (bool, optional) – Whether to save the optimization result (default: False).

minimize(fun, x0, fun_jac=None, fun_hess=None, hessp=None)[source]#

Minimize objective function.

Parameters:

  • fun (callable) – The objective function to be minimized.

  • x0 (array-like) – Initial guess.

  • fun_jac (callable) – Jacobian of fun.

  • fun_hess (callable) – Hessian of fun.

  • hessp (callable)

geomfum.numerics.optimization.result_to_backend_type(result)[source]#

Convert np.array to gs.array within result object.

Module contents#

Numerics Module. This module contains numerical methods and utilities used in Geomfum.