Self-Organizing Neural Network Approach to the Automatic Layout of Business Process Diagrams

• US 20160179344A1
• Filed: 08/13/2015
• Published: 06/23/2016
• Est. Priority Date: 12/22/2014
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Mourra, John
• Granted Patent: US 9,959,518 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 3/04817 using icons graphical or visual programming using iconic symbols G06F8/34
• CPC Class Codes: G06Q 10/067 Enterprise or organisation modelling

Abstract

A method, system, and/or computer program product generates self-organizing layouts of process diagrams. Initial weight vectors are distributed uniformly within boundaries of regions in the process diagram. A spatial input vector is randomly generated within the boundaries of each region. In each region in the process diagram, a closest graphical node is found, and a position of a winning graphical node that is the closest graphical node to the random input vector is adjusted. Positions of all non-immutable graphical objects, w_i, in a topographical neighborhood N(k) of a closest graphical node w_c that can cross a boundary of one or more regions from the multiple regions are adjusted. The spatial input vector is recursively generated, the closest graphical node is recursively located, and the positions of all non-immutable graphical objects, w_i, in the topographical neighborhood N(k) are recursively adjusted until a maximum number of iterations, k_max is reached.

Claims

1. A method for generating self-organizing layouts of process diagrams, the method comprising:

randomly selecting, by one or more processors, initial spatial position vectors, w_i, for all graphical nodes in a process diagram, wherein each graphical node in the process diagram has an initial spatial position vector associated therewith;

distributing, by one or more processors, the initial spatial position vectors, w_i, uniformly within boundaries of multiple regions in the process diagram;

in each region in the process diagram, randomly generating, by one or more processors, a spatial input vector, x, within the boundaries of each region;

in each region in the process diagram, locating, by one or more processors, a closest graphical node, w_c, to a random input vector generated in said each region;

in each region in the process diagram, adjusting, by one or more processors, a position of a winning graphical node that is the closest graphical node to the random input vector generated in said each region, wherein said adjusting the position of the winning graphical node brings the winning graphical node closer to the random input vector;

adjusting, by one or more processors, a weighted vector associated with each immutable closest graphical node in the process diagram, wherein each immutable closest graphical node w, has a fixed location on the process diagram;

adjusting, by one or more processors, positions of all non-immutable graphical objects in a topographical neighborhood N(k) of the closest graphical node w, that can cross a boundary of one or more regions from the multiple regions in the process diagram; and

one or more processors, randomly generating the spatial input vector x, locating the closest graphical node w_c, adjusting the position of the winning graphical node, adjusting the position of the immutable closest graphical node w_c, and adjusting positions of all non-immutable graphical objects in the topographical neighborhood N(k) in a recursive manner until a maximum number of iterations, k_max is reached.

2. The method of claim 1, wherein the closest graphical node w_c is determined by:

minimizing, by one or more processors, a Euclidean distance between x and w_i such that;

∥

x−

w_c∥

=min_i∥

x−

w_i∥

.

3. The method of claim 1, wherein w_c is not immutable, wherein w_c is from a set of non-immutable graphical nodes, w_c, in an interconnected neighborhood N(k), wherein nodes from w_c can cross the regional boundaries in the process diagram, and wherein spatial adjustment of nodes in the process diagram is calculated by one or more processors according to:

Δ

w_i=α

(k)h(n_c, n_i), n_i ∈

N(k)

where α

(k) is an adaption rate, having a value between 0 and 1, that diminishes with every iteration k;

h(n_c, n_i) is a neighborhood function, having a value between 0 and 1, that diminishes as the topographical distance increases between the winning graphical node and other nodes n_i in the neighborhood N(k) and wherein N(k) defines a neighborhood boundary during an iteration k and decreases over a span of the iterations, such that n_i ∈

N(k) refers to every k iteration of output nodes n in the neighborhood N(k).

4. The method of claim 3, further comprising:

determining, by one or more processors, α

(k) h(n_c, n_i) according to;

5. The method of claim 1, further comprising:

recursively randomly generating, by one or more processors, the spatial input vector x, locating the closest graphical node w_c, and adjusting, by one or more processors, the position of the immutable closest graphical node w_c, until a weight adaption rate drops below a predefined threshold.

6. The method of claim 1, further comprising setting boundaries for each region R in the process diagram according to one or more processors implementing an algorithm:

w_ik+1=w_i+Δ

w_i(w_ik+1x, w_ik+1y)

such that;

if w_ik+1x>

x_max^R then w_ik+1x=x_max^R

if w_ik+1x<

x_min^R then w_ik+1x=x_min^R

if w_ik+1y>

y_max^R then w_ik+1y=y_max^R

if w_ik+1y<

y_min^R then w_ik+1y=y_min^R.

7. The method of claim 1, further comprising determining w_c by one or more processors implementing an algorithm:

8. The method of claim 1, further comprising:

characterizing, by one or more processors, N(k) by a radius, r, wherein r denotes a maximum number of connections in a path between two nodes predetermined to be in a same neighborhood, wherein r starts at a pre-selected r_max and is reduced by 1 at a pre-selected k_r interval, until r reaches a pre-selected r_min.