A distance calculation system and method for clustering algorithm

Abstract

A distance calculation system and method for clustering algorithms are proposed. Based on the information distance theory with Coriolis complexity, a unified method for information distance calculation, lossless compressed distance calculation, and lossy compressed distance calculation is designed, and a compressed distance calculation system under lossy conditions is formed. This solves the problem of lossy compressed distance calculation for continuous data such as images and videos, and forms a complete and unified compressed distance calculation method and system applicable to any data type.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence technology, specifically relating to a distance calculation system and method for clustering algorithms. Background Technology

[0002] The current problems with clustering algorithm distance calculation methods and systems include the following:

[0003] Most current clustering methods are feature-based, classifying data by extracting certain features and comparing them. This approach has a limitation: it only captures a limited range of complete data features. Therefore, in recent years, researchers in this field have focused on clustering based on non-feature similarity. This approach aims to provide a unified similarity metric that encompasses all valid clustering distances. Based on this, a theory of absolute information measurement of data based on Komogrove complexity (hereinafter referred to as Komogrove complexity) has been developed. The information distance between two objects X and Y is defined as the minimum computer procedure required to transform X into Y and Y into X.

[0004] However, this distance is incalculable on a Turing computer. An effective solution is to replace this minimal computer program with a universally optimal source code. This transforms the problem of finding the minimal computer program into choosing the optimal compression scheme for the data. Based on this idea, the concept of compression distance was proposed to approximate information distance, implemented through actual compression encoders. The advantage of this method is that it does not require any background knowledge of the data and has proven in recent years to be an applicable data mining method for any parameters and any features.

[0005] However, the information and compression distances in the above studies are based on discrete data, which is not applicable to continuous data such as images and videos. Recent industry research has introduced distortion into the information distance and used lossy source coding rates to express the clustering distance. As with the lossless case, the computability of this value requires efficient lossy compression schemes. Summary of the Invention

[0006] This invention aims to provide a distance calculation method and system for clustering algorithms. Based on the information distance theory of Coriolis complexity, it designs a unified method for information distance calculation, lossless compression distance calculation, and lossy compression distance calculation, and forms a lossy compression distance calculation system. This solves the problem of lossy compression distance calculation for continuous data such as images and videos, and forms a complete and unified compression distance calculation method and system for any data type.

[0007] Specifically, a system for calculating information distance for clustering algorithms includes a binary computer (11), a binary computer (12), and a maximum information distance calculator (13). The binary computer (11) includes a shortest program (1101) and a length calculator (1102) for calculating the Coriolis complexity. The binary computer (12) includes a shortest program (1201) and a length calculator (1202) for calculating the Coriolis complexity. The maximum information distance calculator (13) takes the Coriolis complexity and the Coriolis complexity as input information and calculates the information distance between input object X and input object Y.

[0008] Preferably, the binary computer takes input object X as input and output object Y as output, calculates the shortest program from input object X to output object Y through the shortest program (1101), and calculates the length of the shortest program through the length calculator (1102) to obtain the Coriolis complexity.

[0009] Preferably, the binary computer II takes input object Y as input and output object X as output. It calculates the shortest program from input object Y to output object X through the shortest program II (1201), and calculates the length of the shortest program through the length calculator II (1202) to obtain the Coriolis complexity II.

[0010] Preferably, the first binary computer is a lossless compression length calculator (21) for calculating the lossless compression length of Y; the second binary computer is a lossless compression length calculator (22) for calculating the lossless compression length of X; and the maximum information distance calculator is a lossless compression distance maximum calculator (23) for obtaining the lossless compression distance between input object X and input object Y by inputting the lossless compression length of X and the lossless compression length of Y.

[0011] Preferably, the lossless compression length calculator (21) for Y includes a lossless compressor (2101), a lossless decompressor (2102), a lossless compression length extractor (2103), and a lossless compression length minimum calculator (2104). The lossless compressor (2101) takes the original value of the discrete object Y as input, and after passing through the lossless decompressor (2102), obtains the output discrete object X estimate. The lossless compression length extractor (2103), under the premise that the output discrete object X estimate can be recovered without errors, extracts the lossless compression length of the original value of the input discrete object Y from the lossless compressor (2101) to obtain the compressed data length. The lossless compression length minimum calculator (2104) takes the compressed data length as input and calculates the lossless compression length L(Y|X) of Y.

[0012] Preferably, the lossless compression length calculator (22) for X includes a second lossless compressor (2201), a second lossless decompressor (2202), a second lossless compression length extractor (2203), and a second lossless compression length minimum calculator (2204). The second lossless compressor (2201) takes the original value of the discrete object X as input, and after passing through the second lossless decompressor (2202), obtains the output discrete object Y estimate. The second lossless compression length extractor (2203), under the premise that the output discrete object Y estimate can be recovered without errors, extracts the lossless compression length of the original value of the input discrete object X from the second lossless compressor (2201) to obtain the compressed data length II. The second lossless compression length minimum calculator (2204) takes the compressed data length II as input and calculates the lossless compression length L(X|Y) of X.

[0013] Preferably, the first binary computer is a lossy compression length calculator (31) for calculating the lossy compression length of Y; the second binary computer is a lossy compression length calculator (32) for calculating the lossy compression length of X; the maximum information distance calculator is a lossy compression distance maximum calculator (33), which obtains the lossy compression distance between input object X and input object Y under the premise that the maximum allowable data distortion is D by inputting the lossy compression length of X and the lossy compression length of Y.

[0014] Preferably, the lossy compression length calculator (31) for Y includes a lossy compressor (3101), a lossy decompressor (3102), a lossy compression length extractor (3103), and a lossy compression length minimum calculator (3104). The lossy compressor (3101) takes the original value of the continuous object Y as input, passes it through the lossy decompressor (3102), and obtains the output continuous object X estimate. The lossy compression length extractor (3103) extracts the lossy compression length of the input continuous object Y from the lossy compressor (3101) under the premise that the output continuous object X estimate can be recovered when the maximum allowable data distortion D is allowed, and obtains the compressed data length. The lossy compression length minimum calculator (3104) takes the compressed data length as input and calculates the lossy compression length R_{Y|X}(D) of Y.

[0015] Preferably, the lossy compression length calculator (32) for X includes a lossy compressor (3201), a lossy decompressor (3202), a lossy compression length extractor (3203), and a lossy compression length minimum calculator (3204). The lossy compressor (3201) takes the original value of the continuous object X as input, and after passing through the lossy decompressor (3202), obtains the output continuous object Y estimate. The lossy compression length extractor (3203), under the premise that the output continuous object Y estimate can be recovered when the maximum allowable data distortion D is allowed, extracts the lossy compression length of the input continuous object X from the lossy compressor (3201) to obtain the compressed data length II. The lossy compression length minimum calculator (3204) takes the compressed data length II as input and calculates the lossy compression length R_{X|Y}(D) of X.

[0016] The present invention also provides a method for calculating information distance for clustering algorithms, including a binary computer (11), a binary computer (12), and an information distance maximum calculator (13); the binary computer (11) includes a shortest program (1101) and a length calculator (1102); the binary computer (12) includes a shortest program (1201) and a length calculator (1202); the binary computer (11) calculates the Coriolis complexity, and the binary computer (12) calculates the Coriolis complexity; the information distance maximum calculator (13) uses the Coriolis complexity and the Coriolis complexity as input information to calculate the information distance between input object X and input object Y.

[0017] Preferably, the shortest procedure (1101) calculates the shortest procedure from input object X to output object Y, and the length calculator (1102) calculates the length of the shortest procedure from input object X to output object Y and obtains the Coriolis complexity.

[0018] The second shortest procedure (1201) calculates the shortest procedure from input object Y to output object X. The second length calculator (1202) calculates the length of the shortest procedure from input object Y to output object X and obtains the Coriolis complexity.

[0019] Preferably, the binary computer is a lossless compression length calculator (21) for Y, comprising a lossless compressor (2101), a lossless decompressor (2102), a lossless compression length extractor (2103), and a lossless compression length minimum calculator (2104); the lossless compression length calculator (22) for X comprises a lossless compressor (2201), a lossless decompressor (2202), a lossless compression length extractor (2203), and a lossless compression length minimum calculator (2204); the information distance maximum calculator is a lossless compression distance maximum calculator (23).

[0020] Preferably, the lossless compressor (2101) takes the original value of the discrete object Y as input, and after passing through the lossless decompressor (2102), obtains the output discrete object X estimate; the lossless compression length extractor (2103), under the premise that the output discrete object X estimate can be recovered without error, extracts the lossless compression length of the original value of the input discrete object Y from the lossless compressor (2101) to obtain the compressed data length one; the lossless compression length minimum value calculator (2104) takes the compressed data length one as input and calculates the lossless compression length L(Y|X) of Y;

[0021] The lossless compressor two (2201) takes the original value of discrete object X as input, and after passing through the lossless decompressor two (2202), it obtains the output discrete object Y estimate; the lossless compression length extractor two (2203), under the premise that the output discrete object Y estimate can be recovered without error, extracts the lossless compression length of the original value of the input discrete object X from the lossless compressor two (2201) to obtain the compressed data length two; the lossless compression length minimum value calculator two (2204) takes the compressed data length two as input and calculates the lossless compression length L(X|Y) of X;

[0022] The lossless compression distance maximum calculator (23) uses the lossless compression length of input X and the lossless compression length of input Y as input information to obtain the lossless compression distance between input object X and input object Y.

[0023] Preferably, the first binary computer is a lossy compression length calculator (31) of Y, which includes a lossy compressor (3101), a lossy decompressor (3102), a lossy compression length extractor (3103), and a lossy compression length minimum calculator (3104); the second binary computer is a lossy compression length calculator (32) of X, which includes a lossy compressor (3201), a lossy decompressor (3202), a lossy compression length extractor (3203), and a lossy compression length minimum calculator (3204); the information distance maximum calculator is a lossy compression distance maximum calculator (33).

[0024] Preferably, the lossy compressor (3101) takes the original value of the continuous object Y as input, and after passing through the lossy decompressor (3102), it obtains the estimated value of the output continuous object X; the lossy compression length extractor (3103) extracts the lossy compression length of the input continuous object Y from the lossy compressor (3101) under the premise that the estimated value of the output continuous object X can be recovered when the maximum allowable data distortion D is allowed, and obtains the compressed data length I; the lossy compression length minimum value calculator (3104) takes the compressed data length I as input and calculates the lossy compression length R_{Y|X}(D) of Y;

[0025] The lossy compression length calculator (32) for X: The lossy compressor (3201) takes the original value of the continuous object X as input, passes through the lossy decompressor (3202), and obtains the output continuous object Y estimate; The lossy compression length extractor (3203) extracts the lossy compression length of the input continuous object X from the lossy compressor (3201) under the premise that the output continuous object Y estimate can be recovered when the maximum allowable data distortion D is allowed, and obtains the compressed data length II; The lossy compression length minimum calculator (3204) takes the compressed data length II as input and calculates the lossy compression length R_{X|Y}(D) of X;

[0026] The lossy compression distance maximum calculator (33) obtains the lossy compression distance between input object X and input object Y under the premise that the maximum allowable data distortion is D by inputting the lossy compression length of X and the lossy compression length of Y.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] (1) Based on the absolute information measurement theory of data based on Komogrove complexity, this invention designs a method for calculating the information distance between two objects X and Y. By calculating the distance from X to Y and from Y to X, and calculating the maximum value of the two, the information distance between the two is obtained.

[0029] (2) A method for calculating compression distance under lossless conditions was designed. Compared with the prior art, this method is applicable to the calculation of compression distance based on the information distance theory and has the potential to be extended to the calculation of lossy compression distance.

[0030] (3) Current methods are all lossless compression distance calculation methods for discrete data. This paper proposes and designs a lossy compression distance calculation method, which can be used for clustering distance calculation of continuous data such as images and videos. By introducing the concept of distortion, based on the above calculation method, a lossy compression distance calculation system is designed under the condition of maximum allowable distortion. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of an information distance calculation system and method for clustering algorithms according to the present invention;

[0032] Figure 2 This is a schematic diagram of a compression distance calculation system under non-destructive conditions according to the present invention;

[0033] Figure 3 This is a schematic diagram of a lossless compression length calculator for Y according to the present invention;

[0034] Figure 4 This is a schematic diagram of a lossless compression length calculator for X according to the present invention;

[0035] Figure 5 This is a schematic diagram of a compression distance calculation system and method under lossy conditions according to the present invention;

[0036] Figure 6 This is a schematic diagram of a lossy compression length calculator for Y according to the present invention;

[0037] Figure 7 This is a schematic diagram of a lossy compression length calculator for X according to the present invention;

[0038] Figure 8 This is a schematic diagram of the compression distance calculation system under non-destructive conditions according to Embodiment 1 of the present invention;

[0039] Figure 9 This is a schematic diagram of the XY distance conversion of the compression distance calculation system under non-destructive conditions in Embodiment 1 of the present invention;

[0040] Figure 10 This is a schematic diagram of the X-distance conversion of the compression distance calculation system under non-destructive conditions in Embodiment 1 of the present invention;

[0041] Figure 11 This is a schematic diagram of the Y-distance conversion of the compression distance calculation system under non-destructive conditions in Embodiment 1 of the present invention;

[0042] Figure 12 This is a schematic diagram of the compression distance calculation system under lossy conditions according to Embodiment 2 of the present invention;

[0043] Figure 13 This is a schematic diagram of the XY distance conversion of the compression distance calculation system under lossy conditions in Embodiment 2 of the present invention;

[0044] Figure 14 This is a schematic diagram of the X-distance conversion of the compression distance calculation system under lossy conditions in Embodiment 2 of the present invention;

[0045] Figure 15 This is a schematic diagram of the Y-distance conversion in the lossy compression distance calculation system of Embodiment 2 of the present invention; Detailed Implementation

[0046] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0047] like Figure 1 As shown, an information distance calculation system for clustering algorithms is specifically a method for calculating the information distance between objects X and Y, including binary computer one (11), binary computer two (12), and information distance maximum calculator (13).

[0048] The binary computer (11) includes the shortest program (1101) and the length calculator (1102);

[0049] Binary Computer II (12) includes Shortest Program II (1201) and Length Calculator II (1202);

[0050] Binary computer 1 (11) calculates the Coriolis complexity 1. The input object X is used as the input of binary computer 1 (11), and the output object Y is used as the output of binary computer 1 (11). Binary computer 1 (11) calculates the shortest program from input object X to output object Y through shortest program 1 (1101), and calculates the length of the shortest program through length calculator 1 (1102), and obtains the Coriolis complexity 1, denoted as K(Y|X), which will be used as the distance from input object X to output object Y in the information distance calculation process.

[0051] Binary computer 2 (12) calculates the Coriolis complexity 2. The input object Y is used as the input of binary computer 2 (12), and the output object X is used as the output of binary computer 2 (12). Binary computer 2 (12) calculates the shortest program from input object Y to output object X through shortest program 2 (1201), and calculates the length of the shortest program through length calculator 2 (1202), and obtains the Coriolis complexity 2, labeled as K(X|Y), which will be used as the distance from input object Y to output object X in the information distance calculation process.

[0052] The Coriolis complexity K(Y|X) calculated by binary computer one (11) and the Coriolis complexity K(X|Y) calculated by binary computer two (12) are used together to input the maximum information distance calculator (13) to obtain the information distance between input object X and input object Y, which is marked as ID(X,Y).

[0053] In one or more embodiments, the information distance calculation system for clustering algorithms provided by the present invention can be applied to compressed distance calculation scenarios under lossless conditions. Specifically, as shown in the following examples... Figure 2 As shown, it includes a lossless compression length calculator for Y (21), a lossless compression length calculator for X (22), and a lossless compression distance maximum calculator (23).

[0054] The lossless compression length of Y, calculated by the lossless compression length calculator of Y (21), is denoted as L(Y|X); the lossless compression length of X, calculated by the lossless compression length calculator of X (22), is denoted as L(X|Y). These two lossless compression lengths are input into the lossless compression distance maximum calculator (23) to obtain the lossless compression distance between input object X and input object Y, denoted as CD_II(X,Y).

[0055] In the above embodiment, the lossless compression length calculator (21) of Y calculates the lossless compression length L(Y|X) of Y. Figure 3 As shown, the original value of the input discrete object Y is used as the input of lossless compressor one (2101), and then passes through lossless decompressor one (2102) to obtain the estimated value of the output discrete object X. Under the premise that the estimated value of the output discrete object X can be recovered without errors, the lossless compression length extractor one (2103) extracts the lossless compression length of the original value of the input discrete object Y from the lossless compressor one (2101) to obtain the compressed data length one. Then, the lossless compression length minimum calculator one (2104) is used to calculate the lossless compression length L(Y|X) of Y.

[0056] In the above embodiment, the lossless compression length calculator (22) of X calculates the lossless compression length L(X|Y) of X. Figure 4 As shown, the original value of the input discrete object X is used as the input of lossless compressor two (2201), and then passes through lossless decompressor one (2202) to obtain the estimated value of the output discrete object Y. Under the premise that the estimated value of the output discrete object Y can be recovered without errors, the lossless compression length extractor two (2203) extracts the lossless compression length of the original value of the input discrete object X from lossless compressor two (2201) to obtain the compressed data length two. Then, the lossless compression length minimum calculator two (2204) calculates the lossless compression length L(X|Y) of X.

[0057] In one or more embodiments, the information distance calculation system for clustering algorithms provided by the present invention can be applied to compressed distance calculation scenarios under lossy conditions. Specifically, as shown in the following examples... Figure 5 As shown, it includes a lossy compression length calculator for Y (31), a lossy compression length calculator for X (32), and a lossy compression distance maximum calculator (33).

[0058] Under the premise that the maximum data distortion is allowed to be D, the lossy compression length of Y calculated by the lossy compression length calculator (31) is denoted as R_{Y|X}(D); under the premise that the maximum data distortion is allowed to be D, the lossy compression length of X calculated by the lossy compression length calculator (32) is denoted as R_{X|Y}(D). These two lossy compression lengths are input together into the lossy compression distance maximum calculator (33) to obtain the lossy compression distance between input object X and input object Y under the premise that the maximum data distortion is allowed to be D, denoted as CD_Iy(X,Y).

[0059] In the above embodiments, such as Figure 6 As shown, the lossy compression length calculator (31) calculates the lossy compression length R_{Y|X}(D) of Y under the premise that the maximum allowable data distortion is D. The original value of the input continuous object Y is used as the input of the lossy compressor (3101), and then passes through the lossy decompressor (3102) to obtain the estimated value of the output continuous object X. Under the premise that the estimated value of the output continuous object X can be recovered when the maximum allowable data distortion is D, the lossy compression length extractor (3103) extracts the lossy compression length of the original value of the input continuous object Y from the lossy compressor (3101) to obtain the compressed data length. Then, the lossy compression length minimum calculator (3104) calculates the lossy compression length R_{Y|X}(D) of Y under the premise that the maximum allowable data distortion is D.

[0060] In the above embodiments, such as Figure 7As shown, the lossy compression length calculator (32) calculates the lossy compression length R_{X|Y}(D) of X under the premise that the maximum allowable data distortion is D. The original value of the input continuous object X is used as the input of the lossy compressor two (3201), and then passes through the lossy decompressor two (3202) to obtain the estimated value of the output continuous object Y. Under the premise that the estimated value of the output continuous object Y can be recovered when the maximum allowable data distortion is D, the lossy compression length extractor two (3203) extracts the lossy compression length of the original value of the input continuous object X from the lossy compressor two (3201) to obtain the compressed data length two. Then, it enters the lossy compression length minimum calculator two (3204) to calculate the lossy compression length R_{X|Y}(D) of X under the premise that the maximum allowable data distortion is D.

[0061] This invention also provides a method for calculating information distance in clustering algorithms, such as... Figure 1 As shown, it includes a binary computer one (11), a binary computer two (12), and an information distance maximum calculator (13); the binary computer one (11) includes a shortest program one (1101) and a length calculator one (1102); the binary computer two (12) includes a shortest program two (1201) and a length calculator two (1202); the binary computer one (11) calculates the Coriolis complexity one, and the binary computer two (12) calculates the Coriolis complexity two; the information distance maximum calculator (13) takes the Coriolis complexity one and the Coriolis complexity two as input information and calculates the information distance between input object X and input object Y.

[0062] In the above embodiments, such as Figure 1 As shown, the shortest program one (1101) calculates the shortest program to obtain the output object Y from the input object X, and the length calculator one (1102) calculates the length of the shortest program to obtain the output object Y from the input object X and obtains the Coriolis complexity one;

[0063] The shortest procedure two (1201), as follows: Figure 3 As shown, the shortest program to obtain the output object X from the input object Y is calculated. The length calculator II (1202) calculates the length of the shortest program to obtain the output object X from the input object Y and obtains the Coriolis complexity II.

[0064] In one or more embodiments, the information distance calculation method for clustering algorithms provided by the present invention can be applied to compressed distance calculation scenarios under lossless conditions. For example... Figure 2As shown, the binary computer is a lossless compression length calculator (21) for Y, which includes a lossless compressor (2101), a lossless decompressor (2102), a lossless compression length extractor (2103), and a lossless compression length minimum calculator (2104); the lossless compression length calculator (22) for X includes a lossless compressor (2201), a lossless decompressor (2202), a lossless compression length extractor (2203), and a lossless compression length minimum calculator (2204); the information distance maximum calculator is a lossless compression distance maximum calculator (23).

[0065] In the above embodiments, such as Figure 2-4 As shown, the lossless compressor (2101) takes the original value of the discrete object Y as input, and after passing through the lossless decompressor (2102), it obtains the output discrete object X estimate; the lossless compression length extractor (2103), under the premise that the output discrete object X estimate can be recovered without error, extracts the lossless compression length of the original value of the input discrete object Y from the lossless compressor (2101) to obtain the compressed data length one; the lossless compression length minimum value calculator (2104) takes the compressed data length one as input and calculates the lossless compression length L(Y|X) of Y;

[0066] The lossless compressor two (2201) takes the original value of discrete object X as input, and after passing through the lossless decompressor two (2202), it obtains the output discrete object Y estimate; the lossless compression length extractor two (2203), under the premise that the output discrete object Y estimate can be recovered without error, extracts the lossless compression length of the original value of the input discrete object X from the lossless compressor two (2201) to obtain the compressed data length two; the lossless compression length minimum value calculator two (2204) takes the compressed data length two as input and calculates the lossless compression length L(X|Y) of X;

[0067] The lossless compression distance maximum calculator (23) uses the lossless compression length of input X and the lossless compression length of input Y as input information to obtain the lossless compression distance between input object X and input object Y.

[0068] In one or more embodiments, the information distance calculation method for clustering algorithms provided by the present invention can be applied to compressed distance calculation scenarios under lossy conditions. For example... Figure 5As shown, the first binary computer is a lossy compression length calculator (31) of Y, which includes a lossy compressor (3101), a lossy decompressor (3102), a lossy compression length extractor (3103), and a lossy compression length minimum calculator (3104); the second binary computer is a lossy compression length calculator (32) of X, which includes a lossy compressor (3201), a lossy decompressor (3202), a lossy compression length extractor (3203), and a lossy compression length minimum calculator (3204); the information distance maximum calculator is a lossy compression distance maximum calculator (33).

[0069] In the above embodiments, such as Figure 5-7 As shown, the lossy compressor (3101) takes the original value of the continuous object Y as input, and after passing through the lossy decompressor (3102), it obtains the estimated value of the output continuous object X; the lossy compression length extractor (3103) extracts the lossy compression length of the input continuous object Y from the lossy compressor (3101) under the premise that the estimated value of the output continuous object X can be recovered when the maximum allowable data distortion D is allowed, and obtains the compressed data length I; the lossy compression length minimum value calculator (3104) takes the compressed data length I as input and calculates the lossy compression length R_{Y|X}(D) of Y;

[0070] The lossy compression length calculator (32) for X: The lossy compressor (3201) takes the original value of the continuous object X as input, passes through the lossy decompressor (3202), and obtains the output continuous object Y estimate; The lossy compression length extractor (3203) extracts the lossy compression length of the input continuous object X from the lossy compressor (3201) under the premise that the output continuous object Y estimate can be recovered when the maximum allowable data distortion D is allowed, and obtains the compressed data length II; The lossy compression length minimum calculator (3204) takes the compressed data length II as input and calculates the lossy compression length R_{X|Y}(D) of X;

[0071] The lossy compression distance maximum calculator (33) obtains the lossy compression distance between input object X and input object Y under the premise that the maximum allowable data distortion is D by inputting the lossy compression length of X and the lossy compression length of Y.

[0072] Example 1

[0073] Figure 8 This document describes a specific implementation of a lossless compression distance calculation system for clustering algorithms. The lossless compression distance calculation between X and Y is specifically transformed into three aspects: the lossless compression length of XY, the lossless compression length of X, and the lossless compression length of Y.

[0074] The lossless compression length of XY is calculated by the lossless compression length calculator (51) to obtain the compressed data length L(XY); the lossless compression length of X is calculated by the lossless compression length calculator (52) to obtain the compressed data length L(X); the lossless compression length of Y is calculated by the lossless compression length calculator (53) to obtain the compressed data length L(Y).

[0075] The compressed data length X and the compressed data length Y are input together into the discrete object minimum value calculator (54) to obtain the minimum value, and together with the compressed data length L (XY), they are input into the lossless compression distance adder (55) to obtain the calculated value of the lossless compression distance.

[0076] The XY lossless compression length calculator (51) is implemented as follows: the original value of the discrete object XY is input as the input of the XY lossless compressor (5101), and then passes through the XY lossless decompressor (5102) to obtain the output discrete object XY estimate. Under the premise that the output discrete object XY estimate can be recovered without errors, the lossless compression length of the input discrete object XY original value is extracted from the XY lossless compressor (5101) through the XY lossless compressor (5103). Then, the XY lossless compression length minimum value calculator (5104) is entered to calculate the compressed data length L(XY).

[0077] The X lossless compression length calculator (52) is implemented as follows: the original value of the discrete object X is input as the input of the X lossless compressor (5201), and then passes through the X lossless decompressor (5202) to obtain the output discrete object X estimate. Under the premise that the output discrete object X estimate can be recovered without errors, the lossless compression length of the original value of the input discrete object X is extracted from the X lossless compressor (5201) through the X lossless compression length extractor (5203). Then, the X lossless compression length minimum value calculator (5204) is entered to calculate the compressed data length L(X).

[0078] The lossless compression length calculator (53) for Y is specifically implemented as follows: the original value of the discrete object Y is input as the input of the lossless Y compressor (5301), and then passes through the lossless Y decompressor (5302) to obtain the output value of the discrete object Y. Under the premise that the output value of the discrete object Y can be recovered without errors, the lossless compression length extractor (5303) extracts the lossless compression length of the original value of the input discrete object Y from the lossless Y compressor (5301). Then, the minimum value calculator for the lossless compression length of Y (5304) is used to calculate the length L(Y) of the compressed data.

[0079] Example 2

[0080] Figure 12This document describes a specific implementation of a lossy compression distance calculation system for clustering algorithms. The lossy compression distance calculation between X and Y is specifically transformed into three aspects: the lossy compression length of XY, the lossy compression length of X, and the lossy compression length of Y, all under the premise of allowing a maximum data distortion of D.

[0081] The lossy compression length of XY is calculated by the XY lossy compression length calculator (61) to obtain the compressed data length R_{XY}(D); the lossy compression length of X is calculated by the X lossy compression length calculator (62) to obtain the compressed data length R_{X}(D); the lossy compression length of Y is calculated by the Y lossy compression length calculator (63) to obtain the compressed data length R_{Y}(D).

[0082] The compressed data length X and the compressed data length Y are input together into the discrete object minimum value calculator (64) to obtain the minimum value, and are input together with the compressed data length R_{XY}(D) into the lossy compression distance adder (65) to obtain the calculated value of the lossy compression distance under the premise that the maximum data distortion is allowed to be D.

[0083] The XY lossy compression length calculator (61) is specifically implemented as follows: the original value of the continuous object XY is input as the input of the XY lossy compressor (6101), and then after passing through the XY lossy decompressor (6102), the output continuous object XY estimate is obtained under the premise that the maximum data distortion is allowed to be D.

[0084] Provided that the output continuous object XY estimate can be recovered under the maximum allowable distortion D, the lossy compression length of the original input continuous object XY value is extracted from the XY lossy compressor (6101) by the XY lossy compression length extractor (6103). Then, the XY lossy compression length minimum value calculator (6104) is used to calculate the compressed data length R_{XY}(D).

[0085] The X lossy compression length calculator (62) is specifically implemented as follows: the original value of the continuous object X is input as the input of the X lossy compressor (6201), and then passes through the X lossy decompressor (6202) to obtain the output continuous object X estimate under the premise that the maximum data distortion is allowed to be D.

[0086] Provided that the estimated value of the output continuous object X can be recovered with a maximum allowable data distortion of D, the lossy compression length of the original value of the input continuous object X is extracted from the lossy compressor X (6201) by the lossy compression length extractor X (6203). Then, the minimum value calculator for lossy compression length X (6204) is used to calculate the compressed data length R_{X}(D).

[0087] The Y lossy compression length calculator (63) is specifically implemented as follows: the original value of the continuous object Y is input as the input of the Y lossy compressor (6301), and then after passing through the Y lossy decompressor (6302), the output continuous object Y estimate is obtained under the premise that the maximum data distortion is allowed to be D.

[0088] Provided that the output continuous object Y estimate can be recovered with a maximum allowable data distortion of D, the lossy compression length of the original input continuous object Y is extracted from the lossy Y compressor (6301) by the Y lossy compression length extractor (6303). Then, the Y lossy compression length minimum value calculator (6304) is used to calculate the compressed data length R_{Y}(D).

[0089] The parts of this invention not described in detail are well-known in the art. The above descriptions are merely preferred embodiments of the invention and are not intended to limit the technical solutions of the invention in any way. Any simple modifications, changes in form, and alterations made to the above embodiments based on the technical essence of this invention fall within the protection scope of this invention.

Claims

1. A system for calculating information distance for clustering algorithms, characterized in that, The system includes a binary computer (11), a binary computer (12), and a maximum information distance calculator (13). The binary computer (11) includes a shortest program (1101) and a length calculator (1102) for calculating the Coriolis complexity. The binary computer (12) includes a shortest program (1201) and a length calculator (1202) for calculating the Coriolis complexity. The maximum information distance calculator (13) uses the Coriolis complexity and the Coriolis complexity as input information to calculate the information distance between input object X and input object Y. The first binary computer is a lossy compression length calculator (31) for Y, used to calculate the lossy compression length of Y; the second binary computer is a lossy compression length calculator (32) for X, used to calculate the lossy compression length of X; the maximum information distance calculator is a lossy compression distance maximum calculator (33), which obtains the lossy compression distance between input object X and input object Y under the premise that the maximum allowable data distortion is D by inputting the lossy compression length of X and the lossy compression length of Y; The lossy compression length calculator (31) for Y includes a lossy compressor (3101), a lossy decompressor (3102), a lossy compression length extractor (3103), and a lossy compression length minimum calculator (3104). The lossy compressor (3101) takes the original value of the continuous object Y as input, passes it through the lossy decompressor (3102), and obtains the output continuous object X estimate. The lossy compression length extractor (3103) extracts the lossy compression length of the input continuous object Y from the lossy compressor (3101) under the premise that the output continuous object X estimate can be recovered when the maximum allowable data distortion D is allowed, and obtains the compressed data length. The lossy compression length minimum calculator (3104) takes the compressed data length as input and calculates the lossy compression length of Y. The lossy compression length calculator (32) for X includes a lossy compressor (3201), a lossy decompressor (3202), a lossy compression length extractor (3203), and a lossy compression length minimum calculator (3204). The lossy compressor (3201) takes the original value of the continuous object X as input and passes it through the lossy decompressor (3202) to obtain the estimated value of the output continuous object Y. The lossy compression length extractor (3203) extracts the lossy compression length of the input continuous object X from the lossy compressor (3201) under the premise that the estimated value of the output continuous object Y can be recovered when the maximum allowable data distortion D is allowed, to obtain the compressed data length II. The lossy compression length minimum calculator (3204) takes the compressed data length II as input and calculates the lossy compression length of X.

2. A system for calculating information distance for clustering algorithms as described in claim 1, characterized in that, The binary computer takes input object X as input and output object Y as output. It calculates the shortest program from input object X to output object Y through the shortest program (1101), and calculates the length of the shortest program through the length calculator (1102) to obtain the Coriolis complexity.

3. A system for calculating information distance for clustering algorithms as described in claim 1, characterized in that, The binary computer II takes input object Y as input and output object X as output. It calculates the shortest program from input object Y to output object X through the shortest program II (1201), and calculates the length of the shortest program through the length calculator II (1202) to obtain the Coriolis complexity II.

4. A system for calculating information distance for clustering algorithms as described in claim 1, characterized in that, The first binary computer is a lossless compression length calculator (21) for Y, used to calculate the lossless compression length of Y; the second binary computer is a lossless compression length calculator (22) for X, used to calculate the lossless compression length of X; the information distance maximum calculator is a lossless compression distance maximum calculator (23), which obtains the lossless compression distance between input object X and input object Y by inputting the lossless compression length of X and the lossless compression length of Y.

5. A system for calculating information distance for clustering algorithms as described in claim 4, characterized in that, The lossless compression length calculator (21) for Y includes a lossless compressor (2101), a lossless decompressor (2102), a lossless compression length extractor (2103), and a lossless compression length minimum calculator (2104). The lossless compressor (2101) takes the original value of the discrete object Y as input and passes it through the lossless decompressor (2102) to obtain the output value of the discrete object X. The lossless compression length extractor (2103) extracts the lossless compression length of the original value of the input discrete object Y from the lossless compressor (2101) under the premise that the output value of the discrete object X can be recovered without error, to obtain the compressed data length. The lossless compression length minimum calculator (2104) takes the compressed data length as input and calculates the lossless compression length of Y.

6. A system for calculating information distance for clustering algorithms as described in claim 4, characterized in that, The lossless compression length calculator (22) for X includes a second lossless compressor (2201), a second lossless decompressor (2202), a second lossless compression length extractor (2203), and a second lossless compression length minimum calculator (2204). The second lossless compressor (2201) takes the original value of the discrete object X as input and passes it through the second lossless decompressor (2202) to obtain the output value of the discrete object Y. The second lossless compression length extractor (2203) extracts the lossless compression length of the original value of the input discrete object X from the second lossless compressor (2201) under the premise that the output value of the discrete object Y can be recovered without error, to obtain the compressed data length II. The second lossless compression length minimum calculator (2204) takes the compressed data length II as input and calculates the lossless compression length of X.

7. A method for calculating information distance in clustering algorithms, characterized in that, The system includes a binary computer (11), a binary computer (12), and a maximum information distance calculator (13); the binary computer (11) includes a shortest program (1101) and a length calculator (1102); the binary computer (12) includes a shortest program (1201) and a length calculator (1202); the binary computer (11) calculates the Coriolis complexity I, and the binary computer (12) calculates the Coriolis complexity II; the maximum information distance calculator (13) uses the Coriolis complexity I and the Coriolis complexity II as input information to calculate the information distance between input object X and input object Y; The first binary computer is a lossy compression length calculator (31) for Y, comprising a lossy compressor (3101), a lossy decompressor (3102), a lossy compression length extractor (3103), and a lossy compression length minimum calculator (3104); the second binary computer is a lossy compression length calculator (32) for X, comprising a lossy compressor (3201), a lossy decompressor (3202), a lossy compression length extractor (3203), and a lossy compression length minimum calculator (3204); the maximum information distance calculator is a lossy compression distance maximum calculator (33). The lossy compressor (3101) takes the original value of the continuous object Y as input, and after passing through the lossy decompressor (3102), it obtains the estimated value of the output continuous object X; the lossy compression length extractor (3103) extracts the lossy compression length of the input continuous object Y from the lossy compressor (3101) under the premise that the estimated value of the output continuous object X can be recovered when the maximum allowable data distortion D is allowed, and obtains the compressed data length one; the lossy compression length minimum value calculator (3104) takes the compressed data length one as input and calculates the lossy compression length of Y; The lossy compression length calculator (32) for X: The lossy compressor two (3201) takes the original value of the continuous object X as input, passes through the lossy decompressor two (3202), and obtains the output continuous object Y estimate; The lossy compression length extractor two (3203) extracts the lossy compression length of the input continuous object X from the lossy compressor two (3201) under the premise that the output continuous object Y estimate can be recovered when the maximum allowable data distortion D is allowed, and obtains the compressed data length two; The lossy compression length minimum value calculator two (3204) takes the compressed data length two as input and calculates the lossy compression length of X; The lossy compression distance maximum calculator (33) obtains the lossy compression distance between input object X and input object Y under the premise that the maximum allowable data distortion is D by inputting the lossy compression length of X and the lossy compression length of Y.

8. A method for calculating information distance in a clustering algorithm as described in claim 7, characterized in that, The first shortest procedure (1101) calculates the shortest procedure from input object X to output object Y, and the first length calculator (1102) calculates the length of the shortest procedure from input object X to output object Y and obtains the Coriolis complexity. The second shortest procedure (1201) calculates the shortest procedure from input object Y to output object X, and the second length calculator (1202) calculates the length of the shortest procedure from input object Y to output object X and obtains the Coriolis complexity.

9. A method for calculating information distance in a clustering algorithm as described in claim 7, characterized in that, The binary computer is a lossless compression length calculator (21) for Y, which includes a lossless compressor (2101), a lossless decompressor (2102), a lossless compression length extractor (2103), and a lossless compression length minimum calculator (2104); the lossless compression length calculator (22) for X includes a lossless compressor (2201), a lossless decompressor (2202), a lossless compression length extractor (2203), and a lossless compression length minimum calculator (2204); the information distance maximum calculator is a lossless compression distance maximum calculator (23).

10. A method for calculating information distance in a clustering algorithm as described in claim 9, characterized in that, The lossless compressor (2101) takes the original value of discrete object Y as input, and after passing through the lossless decompressor (2102), it obtains the output discrete object X estimate; the lossless compression length extractor (2103) extracts the lossless compression length of the original value of the input discrete object Y from the lossless compressor (2101) under the premise that the output discrete object X estimate can be recovered without error, and obtains the compressed data length one; the lossless compression length minimum value calculator (2104) takes the compressed data length one as input and calculates the lossless compression length of Y; The lossless compressor two (2201) takes the original value of discrete object X as input, and after passing through the lossless decompressor two (2202), it obtains the output discrete object Y estimate; the lossless compression length extractor two (2203) extracts the lossless compression length of the original value of the input discrete object X from the lossless compressor two (2201) under the premise that the output discrete object Y estimate can be recovered without error, and obtains the compressed data length two; the lossless compression length minimum value calculator two (2204) takes the compressed data length two as input and calculates the lossless compression length of X; The lossless compression distance maximum calculator (23) uses the lossless compression length of input X and the lossless compression length of input Y as input information to obtain the lossless compression distance between input object X and input object Y.

Images