Multidimensional data visualization apparatus, method, and program

Patent metadata

Field	Value
Publication number	US-2022012262-A1
Application number	US-201917282182-A
Country	US
Kind code	A1
Filing date	Sep 20, 2019
Priority date	Oct 4, 2018
Publication date	Jan 13, 2022
Grant date	—

Abstract

An embodiment of the present invention is provided with a projective transform model including a plurality of nodes and a projection table, the plurality of nodes each holding a reference vector having a dimension corresponding to the dimension of multi-dimensional data. The projection table indicates the correspondence relation between the number of each node and a coordinate in a two-dimensional space as a projection target of the reference vector held by the node. First in a learning phase, multi-dimensional input data of a positive example and a negative example is acquired, the amplitude characteristic amounts thereof are calculated, and this amplitude characteristic amount data is learned as the reference vectors of the nodes for each sample. Subsequently, the Euclidean distance between coordinates when the nodes learned based on the amplitude characteristic amount data of the positive example and the nodes learned based on the amplitude characteristic amount data of the negative example are projected into the two-dimensional space in accordance with the projection table is calculated, and coordinates in the projection table are updated so that the calculated Euclidean distance becomes equal to or larger than a threshold value.

First claim

1 . A multi-dimensional data visualization device provided with a projective transform model including a plurality of nodes and a table, the plurality of nodes each holding a reference vector having a dimension identical to a dimension of input multi-dimensional data, the table storing information indicating a correspondence relation between each node and a coordinate in a projection target space, the multi-dimensional data visualization device comprising:
a processor; and a storage medium having computer program instructions stored thereon, when executed by the processor, perform to: acquire first and second multi-dimensional data for learning and extract characteristic amounts from the acquired first and second multi-dimensional data, respectively; update the reference vector of each node based on the extracted characteristic amounts of the first and second multi-dimensional data; and calculate, based on the characteristic amounts of the first and second multi-dimensional data, a distance between each pair of coordinates when the updated nodes are projected into the projection target space in accordance with the table and update the table so that the distance is equal to or larger than a threshold value.

2 . The multi-dimensional data visualization device according to claim 1 , wherein the projective transform model includes, as the table, a table storing information indicating a correspondence relation between each node and a coordinate in a space having a dimension lower than the dimension of the multi-dimensional data.

3 . The multi-dimensional data visualization device according to claim 1 , wherein the projective transform model is established by applying a self-organizing map.

4 . The multi-dimensional data visualization device according to claim 1 wherein the computer program instructions further perform to to acquire third and fourth multi-dimensional data as visualization targets and extract characteristic amounts from the acquired third and fourth multi-dimensional data, respectively; and input the extracted characteristic amounts of the third and fourth multi-dimensional data to the projective transform model after the nodes and the table are updated and generate and output display data for displaying a list of coordinates in the projection target space based on coordinates output from the projective transform model and corresponding to the third and fourth multi-dimensional data in the projection target space.

5 . A multi-dimensional data visualization method executed by a device provided with a projective transform model including a plurality of nodes and a table, the plurality of nodes each holding a reference vector having a dimension identical to a dimension of input multi-dimensional data, the table storing information indicating a correspondence relation between each node and a coordinate in a projection target space, the multi-dimensional data visualization method comprising:
a step of acquiring first and second multi-dimensional data for learning and extracting characteristic amounts from the acquired first and second multi-dimensional data, respectively; a step of updating the reference vector of each node based on the extracted characteristic amounts of the first and second multi-dimensional data; and a step of calculating, based on the characteristic amounts of the first and second multi-dimensional data, a distance between each pair of coordinates when the updated nodes are projected into the projection target space in accordance with the table and updating the table so that the distance is equal to or larger than a threshold value.

6 . The multi-dimensional data visualization method according to claim 5 , further comprising:
a step of acquiring third and fourth multi-dimensional data as visualization targets and extracting characteristic amounts from the acquired third and fourth multi-dimensional data, respectively; a step of inputting the extracted characteristic amounts of the third and fourth multi-dimensional data to the projective transform model after the nodes and the table are updated, and generating and outputting display data for displaying a list of coordinates in the projection target space based on coordinates output from the projective transform model and corresponding to the third and fourth multi-dimensional data in the projection target space.

7 . A non-transitory computer-readable medium having computer-executable instructions that, upon execution of the instructions by a processor of a computer, cause the computer to function as the multi-dimensional data visualization device according to claim 1 .

Assignees

• Nippon Telegraph & Telephone

Inventors

• Isezaki Takashi
• Aoki Ryosuke
• Watanabe Tomoki
• Yamada Tomohiro

Classifications

• G16H10/60

  for patient-specific data, e.g. for electronic patient records · CPC title

• G16H50/20

  for computer-aided diagnosis, e.g. based on medical expert systems · CPC title

• A61B5/7253

  characterised by using transforms · CPC title

• A61B5/742

  using visual displays (displays for heart-related electrical signals, e.g. ECG, A61B5/339) · CPC title

• A61B5/7264

  Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems · CPC title