Helper Module for Deep Learning.

Simple Graph Structure on which we want to perform a topological tree (no cycle).

The algorithm is based on the R.E. Tarjanlinear linear optimization (O(N+A)).

Attributes

__init__()[source]¶

Create a Graph

add_graph(graph)[source]¶

Method to add a Graph in the Graph.

Parameters

graph: Graph (mandatory)

the graph to insert.

add_link(from_node, to_node)[source]¶

Method to add an edge between two GraphNodes of the Graph.

Parameters

from_node: GraphNode (mandatory)

node link representation of the form ‘<node>.<control>’.

to_node: GraphNode (mandatory)

node link representation of the form ‘<node>.<control>’.

add_node(node)[source]¶

Method to add a GraphNode in the Graph.

Parameters

node: GraphNode (mandatory)

the node to insert.

adjacency_matrix()[source]¶

Compute the graph adjacency matrix.

Returns

adjacency_matrix: array

Graph adjacency matrix.

node_names: list

Ordered node names.

available_nodes()[source]¶

List the nodes that have no incoming link.

find_node(node_name)[source]¶

Method to find a GraphNode in the Graph.

Parameters

node_name: str (mandatory)

the name of the desired node.

layout(scale=1.0)[source]¶

Position nodes using the eigenvectors of the graph Laplacian.

Parameters

scale: float (optional, default 1.0)

Scale factor for positions. The nodes are positioned in a box of size [0, scale] x [0, scale].

Returns

pos: dict

A dictionary of positions keyed by node.

remove_node(node_name)[source]¶

Method to remove a GraphNode from the Graph.

Parameters

node: string (mandatory)

the name of the node to remove.

topological_sort()[source]¶

Perform the topological sort: find an order in which all the nodes can be taken.

Step 1: Identify nodes that have no incoming link (nnil). Step 2: Loop until there are nnil a) Delete the current nodes c_nnil of in-degree 0. b) Place it in the output. c) Remove all its outgoing links from the graph. d) If the node has in-degree 0, add the node to nnil. Step 3: Assert that there is no loop in the graph.

Returns

output: list of tuple

a list of ordered nodes with a tuple element containing the node name and the node meta element.

Simple Graph Node Structure.

Attributes

__init__(name, meta)[source]¶

Create a Graph Node.

Parameters

name: str (mandatory)

the name of the node.

meta: object

an python object to store in the node.

add_link_from(node)[source]¶

Method to add a Predecessor.

Parameters

node: GraphNode (mandatory)

the predecessor node.

add_link_to(node)[source]¶

Method to add a Successor.

Parameters

node: GraphNode (mandatory)

the successor node.

remove_link_from(node)[source]¶

Method to remove a Predecessor.

Parameters

node: GraphNode (mandatory)

the predecessor node.

remove_link_to(node)[source]¶

Method to remove a Successor.

Parameters

node: GraphNode (mandatory)

the successor node.