ExVrp.Neighbourhood (ExVrp v0.4.2)

Copy Markdown View Source

Computes granular neighbourhood for local search.

Uses Nx for efficient tensor operations, directly porting PyVRP's compute_neighbours algorithm from pyvrp/search/neighbourhood.py.

Example

# Get neighbours for a problem
{:ok, problem_data} = ExVrp.Native.create_problem_data(model)
params = ExVrp.NeighbourhoodParams.new(num_neighbours: 40)
neighbours = ExVrp.Neighbourhood.compute_neighbours(problem_data, params)

# neighbours is a list of lists:
# - neighbours[0..num_depots-1] are empty (depots have no neighbours)
# - neighbours[i] for clients contains the k nearest client indices

Summary

Functions

compute_neighbours(problem_data, params \\ %NeighbourhoodParams{})

Computes neighbours for each location.

Functions

compute_neighbours(problem_data, params \\ %NeighbourhoodParams{})

@spec compute_neighbours(reference(), ExVrp.NeighbourhoodParams.t()) :: [[integer()]]

Computes neighbours for each location.

Returns list of lists: neighbours[location] = [neighbour_indices...] Depots get empty lists.

Parameters

  • problem_data - Reference to ProblemData resource
  • params - NeighbourhoodParams configuration

Returns

A list of lists where:

  • The first num_depots entries are empty lists
  • Each client entry contains the indices of its k nearest neighbours