Graph Construction API

generate_elastix_difficulty(path_elastix_difficulty::String, t_range, heuristic::Function)

Generates an elastix difficulty file based on the given heuristic.

Arguments

• path_elastix_difficulty::String: output file
• t_range: list or range of time points to compute the difficulty
• heuristic: a heuristic function that evaluates "distance" betwen two frames. The function will be given t1, and t2 as input, so be sure its other parameters have been initialized correctly. It is assumed that the function outputs floating-point values.

generate_elastix_difficulty(param_path::Dict, t_range, heuristic::Function)

Generates an elastix difficulty file based on the given heuristic.

Arguments

• param_path::Dict: Dictionary of paths containing path_elastix_difficulty key to the path of the elastix difficulty output file
• t_range: list or range of time points to compute the difficulty
• heuristic: a heuristic function that evaluates "distance" betwen two frames. The function will be given t1, and t2 as input, so be sure its other parameters have been initialized correctly. It is assumed that the function outputs floating-point values.

load_graph(elx_difficulty::String; func=nothing, difficulty_importance::Real=0.05)

Loads an adjacency matrix from a file and stores it as a graph. You can specify a function to apply to each weight. This is often helpful in cases where the heuristic obeys the triangle inequality, to avoid the function from generating a star graph and mostly ignoring the heuristic. By default, the function is x->x^1.05, which allows the algorithm to split especially difficult registration problems into many steps. If this results in too many failed registrations, try increasnig the difficulty_importance parameter; conversely, if there is too much error accumulation over long chains of registrations, try decreasing it.

Arguments

• elx_difficulty::String: a path to a text file containing a list of frames and an adjacency matrix.
• func (optional): a function to apply to each element of the adjacency matrix
• difficulty_importance (optional, default 0.05): if func is not provided, it will be set to x->x^(1+difficulty_importance)

Returns a graph graph::SimpleWeightedGraph storing the adjacency matrix.

make_voting_subgraph(graph::SimpleWeightedGraph, degree::Integer)

Makes a subgraph of a graph, where each node is connected to their degree closest neighbors. Returns the subgraph, and an array of nodes that were disconnected from the rest of the nodes.

output_graph(subgraph::SimpleWeightedDiGraph, outfile::String; max_fixed_t::Int=0)

Outputs subgraph::SimpleWeightedDiGraph to an output file outfile containing a list of edges in subgraph. Can set max_fixed_t::Int parameter if a dataset-alignment registration is being done.

to_dict(graph::SimpleWeightedGraph)

Converts a graph::SimpleWeightedGraph into an adjacency dictionary node => Dict({neighbor1 => weight1, neighbor2=> weight2, ...})

remove_frame(graph::SimpleWeightedGraph, frame::Integer)

s Given a graph graph::SimpleWeightedGraph and a problematic frame::Integer, deletes the frame from the graph without changing frame indices.

update_graph(reg_quality_arr::Array{String,1}, graph::SimpleWeightedGraph, metric::String; metric_tfm=identity)

Recomputes the difficulty graph based on registration quality data. Returns a new graph where difficulties of registration problems are scaled by the quality metric.

Arguments

• reg_quality_arr::Array{String,1} is an array of paths to files containing registration quality data
• graph::SimpleWeightedGraph is the difficulty graph to be updated
• metric::String is which quality metric to use to update the graph

Optional keyword arguments

• metric_tfm: Function to apply to each metric value. Default identity.

remove_previous_registrations(previous_problems, subgraph::SimpleWeightedDiGraph)

Removes previous registrations from the subgraph.

Arguments:

• previous_problems: list of registration problems
• subgraph::SimpleWeightedDiGraph: current subgraph

optimize_subgraph(graph::SimpleWeightedGraph)

Finds the minimum node of a graph, which has the smallest average shortest path length to each other node in that graph. Unconnected nodes are counted as having a path length equal to the highest edge weight.

Parameters

• graph::SimpleWeightedGraph: a graph

Returns:

• min_node::Integer: the minimum node
• subgraph::SimpleWeightedDiGraph: an unweighted graph whose edges consist of shortest paths from the minimum node of the original graph to other nodes.
• maximum_problem_chain::Integer: the number of vertices in the longest chain of registration problems in the subgraph.

This graph is directed, and only has edges going away from the minimum node.