CN102222233B - Method for realizing data classification of simulation-field computer on basis of magnetic field - Google Patents

Abstract

The invention discloses a method for realizing data classification of a simulation-field computer on the basis of a magnetic field. The method comprises the steps of: selecting M1*M2 points on a magnetic conducting medium film to arrange probes vertically; from the arranged probes, selecting k2 probes as magnetic field construction input points, selecting k1 probes as data input points, and selecting one probe as a data output point; generating n2 k2-dimensional constructing data Oj(k2) at random, wherein j is equal to 1, 2L Ln2, and each constructing data can construct one simulation-field computer; respectively selecting n1 data from each type of data of Qi(k1) as input data; determining the density of current input to the magnetic field construction input points by each dimensional value of the constructing data, determining the density of current input to the data input points by each dimensional value in the input data, and acquiring a magnetic-field intensity value or magnetic-induction intensity value from the data output points as an output value; and carrying out verification on the constructing data, and selecting the optimal constructing data to acquire a simulation-field computer and further realize data classification. Due to adoption of the method, the speed of data classification is high and the accuracy is high.

Description

A kind ofly based on magnetic field, realize simulation yard computer data sorting technique

Technical field

The present invention relates to a kind of data classification method, particularly a kind ofly based on magnetic field, realize simulation yard computer data sorting technique.

Background technology

Pattern-recognition is the mankind's a primary mental ability, and in daily life, people replace the brainwork of mankind's part through conventional " pattern-recognition ".Pattern-recognition refers to be processed and analyzes characterizing the various forms of information of things or phenomenon, with the process that things or phenomenon are described, recognize, are classified and explain.In the process of pattern-recognition, conventionally will classify to a large amount of data, the speed of Data classification and accuracy directly affect the Efficiency and accuracy of pattern-recognition.

At present, conventionally adopting the microprocessor based on silicon materials is basic digital machine, and the mass data of obtaining in mode identification procedure is classified.But, the theoretic light transmission speed limit of Moore's Law and microelectric technique, the bottleneck that quantum-mechanical uncertainty principle and the second law of thermodynamics etc. are brought, makes that digital machine is slow in the process speed of carrying out Data classification, efficiency is low.Digital machine is in the two dimension in the face of producing in pattern-recognition and data more than two dimension simultaneously, and it can only carry out serial processing to every one dimension of data, and its classification effectiveness is low.

Summary of the invention

The object of the invention is to propose a kind of data classification method of realizing simulation yard computing machine based on magnetic field, the method can be carried out parallel processing to each dimension in multidimensional data simultaneously, and adopt the simulation yard computing machine of realizing based on magnetic field, non-linear classification is strong, can realize complex data is carried out to high speed classifying exactly.

Object of the present invention is mainly achieved through the following technical solutions:

Based on magnetic field, realize a data classification method for simulation yard computing machine, for Ψ the k that mode identification procedure is obtained ₁dimension data Q _i[k ₁] classify, wherein, i=1,2 ... Ψ, and Q _i[k ₁] in comprise A type data; Concrete steps are:

Step 1, on magnetic conductive media film, choose M ₁* M ₂individual point vertically arranges probe; From set probe, choose k ₂individual probe, as building magnetic field input point, is chosen k ₁individual probe, as data input point, is chosen a probe as data output point; Wherein, building magnetic field input point, data input point and data output point is mutually different point;

Step 2, generate n at random ₂individual for building the k in magnetic field ₂dimension builds data O _j[k ₂], j=1,2 ... n ₂, each builds data can build a simulation yard computing machine; From Q _i[k ₁] each categorical data in choose respectively n ₁individual data are as input data; The every dimension value that builds data determines to the density that builds magnetic field input point input current, and the determine density of directional data input point input current of the every dimension value in input data is obtained field strength values or magnetic induction density value as output valve from data output point;

Step 3, to building data, verify, its concrete proof procedure is: for each, build data, set output area corresponding to each type input data; By being input in data input point according to the electric current of each input data acquisition, from data output point, obtain one group of output valve; Output valve is respectively organized in utilization and output area calculating respectively builds error corresponding to data, chooses the optimum data that build set up simulation yard computing machine to Q according to described error _i[k ₁] classify.

The process of checking of the present invention is:

Step 201, set output area corresponding to each type input data;

Step 202, from O _j[k ₂] in choose one not traversal structure data as the density value of electric current, choose k ₂individual electric current, and by described k ₂individual electric current is input to k ₂in the input point of individual structure magnetic field, and by the structure data markers of obtaining, be the structure data that traveled through;

Step 203, for each input data, utilize its density value as electric current to obtain k ₁individual electric current, and by described k ₁individual electric current is input to k ₁in individual data input point;

The simulation yard computing machine that step 204, judgement utilize structure data that this circulation chooses to set up, for each input corresponding output valve of data and output area, calculate error corresponding to structure data, judge whether described error meets pre-conditioned, if so, using building data in this circulation, as optimum structure data, set up simulation yard computing machine to Q _i[k ₁] classify, otherwise, return to step 202.

The detailed process of step 2 of the present invention and step 3 is:

Step 301, maximum iteration time is set is Maxiter, and sets the corresponding output area of each type in A type of data, with seasonal iterations iter=0;

Step 302, utilize particle cluster algorithm to generate at random n ₂individual for building the structure data O in magnetic field _j[k ₂], j=1,2 ... n ₂, each dimension that builds data is k ₂, n ₂individual structure data are different between two, and each builds the corresponding margin of error B of data _j, make B _j=0; From Q _i[k ₁] each categorical data in choose respectively n ₁individual data are as input data;

Step 303, from n ₂individual O _j[k ₂] in choose a k who is not traversed ₂dimension builds data, is designated as O _η[k ₂], by O _η[k ₂] middle k ₂each dimension value of dimension data is as the density value of electric current, by k ₂individual electric current is input to respectively k ₂in the input point of individual structure magnetic field, now by described O _η[k ₂] be labeled as the structure data that traveled through;

Step 304, from A * n ₁in individual input data, choose a k who is not traversed ₁dimension input data, are designated as Q _μ[k ₁], by Q _μ[k ₁] middle k ₁each dimension value of dimension data is as the density value of electric current, by k ₁individual electric current is input to respectively k ₁in individual data input point, now by described Q _μ[k ₁] be labeled as the input data that traveled through; Store simultaneously and on data output point, obtain field strength values or magnetic induction density value as output valve;

Step 305, judge above-mentioned A * n ₁whether individual input data are all the input data that traveled through, and if so, enter step 306, otherwise, return to step 304;

Step 306, judgement said n ₂whether individual structure data are all the structure data that traveled through, and if so, enter step 307, otherwise, return to step 303;

Step 307, obtain this circulation in n ₂the corresponding n of individual structure data ₂group output valve, wherein every group has A * n ₁individual output valve, Yi Zuwei unit judges that each output valve is whether in the determined output area of its corresponding input data type, if so, B _jconstant, otherwise make B _jequal B _jadd difference DELTA, wherein said difference DELTA is: in the corresponding output area of output valve and the difference of the immediate value of output valve and output valve square;

Step 308, choose margin of error B _jone of middle minimum, is designated as B _min, and judge B _minwhether be less than default ξ, if enter step 310; Otherwise, enter step 309;

Step 309, judge whether iterations iter is less than maximum iteration time Maxiter, if so, make iterations iter add 1, according to B _jutilize particle cluster algorithm to n ₂individual structure data O _j[k ₂] upgrade, enter step 303, if not, step 310 entered;

Step 310, general be B now _mincorresponding structure data build data as optimum, utilize the optimum data that build to choose k as the density value of electric current ₂individual electric current, by k ₂individual electric current is input to k ₂in the input point of individual structure magnetic field, successively by Q _i[k ₁] from k ₁in individual data input point, input, and according to the residing scope of output valve, the data of input are classified.

Beneficial effect

The present invention is based on magnetic field and utilize the data of many group mode identifications to train a plurality of structure data, obtain the optimum data that build, the simulation yard computing machine that utilizes this optimum to build data realization can be classified to data exactly.

Secondly, when processing multidimensional data, by choosing the number of data input point on simulation yard computing machine, and on each point, input the corresponding electric current of each dimension value, can realize each dimension of multidimensional data is processed simultaneously.

Again, the present invention is based on the simulation yard computing machine that realize in magnetic field, the non-linear effect of data input point and data output point is good, therefore can realize complex data is comparatively processed; The speed that simultaneously forms magnetic field is fast, adopts data processing speed of the present invention fast.

Accompanying drawing explanation

Fig. 1 is the schematic diagram of simulation yard computing machine of the present invention.

Fig. 2 is calculating model for magnetic field schematic diagram of the present invention.

Fig. 3 be one embodiment of the invention to build that data verify process flow diagram.

Fig. 4 is that another two embodiment of the present invention carry out the process flow diagram of sorting technique to data.

Embodiment

Below in conjunction with accompanying drawing, the present invention is further elaborated.

The present invention is based on magnetic field and realize simulation yard computer data sorting technique, for Ψ the k that mode identification procedure reed is got ₁dimension data Q _i[k ₁] classify, wherein, i=1,2 ... Ψ, and Q _i[k ₁] in comprise A type data.

Step 1, on magnetic conductive media film, choose M ₁* M ₂individual point vertically arranges probe; As shown in Figure 1, preferably choose M ₁=M ₂, and the distance between every adjacent 2 is equal, M ₁* M ₂individual point has formed a rectangular array.From the probe of described setting, choose k ₂individual probe, as building magnetic field input point, is chosen k ₁individual probe, as data input point, is chosen a probe as data output point; Wherein, building magnetic field input point, data input point and data output point is mutually different point.

Because probe is arranged on magnetic conductive media film, while therefore having electric current input on a certain probe or certain a plurality of probe, on magnetic conductive media film, form magnetic field, on other probe, there is the output of magnetic field intensity or magnetic induction density.

The present invention chooses k ₂individual probe is as building magnetic field input point, when the data of classifying when needs are comprised of polytype multidimensional data, and k ₂choose greatlyr, the data of classifying when needs are one group of comparatively simple data, when the data type comprising and dimension are all less, and k ₂choose less.For example in the present embodiment, establish k ₁equal 4, A and equal 5, the data that obtain in mode identification procedure are to comprise 4 dimension data of 5 types, by k ₂elect 4 as.Due to Q _i[k ₁] be k ₁dimension data, has represented respectively the k of things ₁individual feature, therefore chooses k ₁individual probe is as data input point, and the density value using each dimension data as electric current is respectively inputted the corresponding electric current of density value of each electric current from data input point.

For example, establish that in the present embodiment, to export data on data output point be magnetic field intensity; When the input current on building magnetic field input point is determined, the electric current of now inputting in data input point can affect the magnetic field intensity of exporting on data output point, between the magnetic field intensity that makes to export on the electric current inputted in data input point and data output point, there is a nonlinear relationship, and then can be according to the corresponding type of output judgement input data on data output point.

The proof of above-mentioned nonlinear relationship is as follows:

Suppose that electric current I is a wireless long straight conductor that passes magnetic conductive media film, sense of current is perpendicular to magnetic conductive media film outside (being defined as forward).As shown in Figure 2, the circle in figure represents the magnetic line of force.The amplitude H that gets magnetic field intensity is output quantity,

direction is the tangential direction of the magnetic line of force,

it is the distance between input point and output point.

As shown in Fig. 2 (a), now on magnetic conductive media film, there are electric current input, a magnetic field intensity for output point

h wherein _x=-H sin α, H _y=H cos α, H is the magnetic field intensity that input current produces on output point, α is the angle of input point and output point line and x axle.

As shown in Fig. 2 (b), now on magnetic conductive media film, there are two electric current inputs, the magnetic field intensitys of output point h wherein _x=-H ₁sin α ₁-H ₂sin α ₂, H _y=H ₁cos α ₁+ H ₂cos α ₂, H ₁be the magnetic field intensity that the first input current produces on output point, H ₂be the magnetic field intensity that the second input current produces on output point, α ₁the angle of first electric current input point and output point line and x axle, α ₂it is the angle of second electric current input point and output point line and x axle.

Suppose that film has m electric current input point (forward; v _i=0, negative sense v _i=1), 1 output point, the magnetic field intensity of output point is:

$\mathrm{Hsum}=\sqrt{{H_x}^2+{H_y}^2}=\frac{1}{2\mathrm{\π}}\sqrt{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{m}{\mathrm{\Σ}}}{(-1)}^{v_i+v_j}{I}_i{I}_j\frac{\cos ({\mathrm{\α}}_i-{\mathrm{\α}}_j)}{{\mathrm{\ρ}}_i{\mathrm{\ρ}}_j}}$

Wherein, α _irepresent the angle between i electric current input point and output point line and x axle, α _jrepresent the angle between j electric current input point and output point line and x axle; ρ _irepresent the distance between i electric current input point and output point, ρ _jdistance between j electric current input point and output point.

This shows, between the magnetic field intensity of output point and the electric current of input point, exist nonlinear relation.

Step 2, generate n at random ₂individual for building the k in magnetic field ₂dimension builds data O _j[k ₂], j=1,2 ... n ₂, each builds data can build a simulation yard computing machine; From Q _i[k ₁] each categorical data in choose respectively n ₁individual data are as input data; The element value that builds data determines to the density that builds magnetic field input point input current, and the determine density of directional data input point input current of the element value in input data is obtained field strength values or magnetic induction density value as output valve from data output point.

Step 3, to building data, verify, its concrete proof procedure is: for each, build data, the electric current of each input data acquisition is input in data input point, from data output point, obtain one group of output valve; Utilize and respectively organize output valve and error corresponding to output area calculating structure data, according to described error, choose the optimum data that build and set up simulation yard computing machine, to Q _i[k ₁] classify.

In one embodiment, as shown in Figure 3, the detailed process that structure data are verified is:

Step 201, set output area corresponding to each type input data;

Step 202, from O _j[k ₂] in choose structure data and choose k as the density value of electric current ₂individual electric current, and by described k ₂individual electric current is input to k ₂in the input point of individual structure magnetic field;

Step 203, for each input data, utilize its density value as electric current to obtain k ₁individual electric current, and by described k ₁individual electric current is input to k ₁in individual data input point;

The simulation yard computing machine that step 204, judgement utilize structure data that this circulation chooses to set up, for each input corresponding output valve of data and output area, calculate error corresponding to structure data, judge described error be enough meet pre-conditioned, if, using building data in this circulation, as optimum structure data, set up simulation yard computing machine, to Q _i[k ₁] classify, otherwise, return to step 202.

In another embodiment, as shown in Figure 4, the present invention can set up in the following way and to the data detailed process of classifying be:

Steps A 1, on magnetic conductive media film, choose M ₁* M ₂individual point vertically arranges probe; From set probe, choose k ₂individual probe, as building magnetic field input point, is chosen k ₁individual probe, as data input point, is chosen a probe as data output point; Wherein, building magnetic field input point, data input point and data output point is mutually different point.

Steps A 2, maximum iteration time is set is Maxiter, and sets the corresponding output area [f of each type in A type of data _a, f _a+1], wherein, a=1,2 ... 5.For example output area corresponding to the first data type is [f ₁, f ₂], output area corresponding to the second data type is [f ₂, f ₃], output area corresponding to the third categorical data type is [f ₃, f ₄] etc., with seasonal iterations iter=0.

Steps A 3, utilize particle cluster algorithm to generate at random n ₂individual for building the structure data in magnetic field, O _j[k ₂], j=1,2 ... n ₂, wherein each dimension that builds data is k ₂dimension, n ₂individual structure data are different between two, and each builds the corresponding margin of error B of data _j, make B _j=0.

The particle cluster algorithm using in this step is comparatively proven technique of recent of developing, and it is from RANDOM SOLUTION, by iteration, finds optimum solution, and in the present invention, optimized variable correspondence the structure data that build magnetic field.Particle cluster algorithm particle value is new formula more:

v(t+1)＝w*v(t)+c ₁*r1*(pb(t)-x(t))+c ₂*r2*(pg(t)-x(t))

x(t+1)＝x(t)+v(t+1)

The t here just refers to the number of times of iteration.V refers to the speed of particle, and x refers to the position of particle.PSO algorithm parameter is selected

C ₁, c ₂it is accelerator coefficient (or claiming the study factor), regulate respectively to the maximum step-length of the best particle of the overall situation and the flight of individual preferably particle direction, if too little, particle may wide region, if too greatly can cause, to target area, fly to suddenly, or fly over target area.Suitable c ₁, c ₂can convergence speedup and be difficult for being absorbed in local optimum.Work as c ₁≈ c ₂time, the work efficiency of particle is the highest, conventionally gets c ₁=c ₂=2.

With inertia weight w, control the impact on present speed of speed above, larger w can strengthen the ability of searching optimum of PSO, and less w can strengthen local search ability.Actual result shows, w has speed of convergence faster between [0.8,1.2].

Iterations: in order to obtain a good solution, needed iterations also depends on particular problem.Iteration very little may make algorithm precocious, and too many iterations can increase the burden of calculating.

Population Size: the population of large quantity can be in iteration each time can search volume in larger region, also will increase the calculated amount of algorithm, and the performance that reduces parallel random search simultaneously.Research experience shows, 10-30 particle is best.But still depend on the problem that specifically will solve, in the solving of this problem, experimental result shows, gets 15 particle best results.

Search volume (Size of Neighborhood): neighborhood is less, exchanges lesser, restrains slowlyer, but its convergence more can be found optimum solution reliably, is not easy to be absorbed in local extremum simultaneously.

The structure data O that steps A 4, utilization are not traversed _η[k ₂] as the density value of electric current, choose k ₂individual electric current, by described k ₂individual electric current is input to k ₂in the input point of individual structure magnetic field, now by O _η[k ₂] be labeled as the structure data that traveled through, η ∈ { 1,2 ... n ₂.

Owing to cannot determining that each builds the size of required input electric current in the input point of magnetic field, could classify accurately to input data, therefore utilize the random n generating of particle cluster algorithm ₂individual structure data, and the Ψ that mode identification procedure is obtained k who comprises A type ₁dimension data, chooses respectively n for each type ₁the data of individual known type, obtain A * n ₁data are trained (study) as input data to building data, to obtain optimum structure data, detailed process is as follows.

The input data Q that steps A 5, utilization are not traversed _μ[k ₁] as the density value of electric current, choose k ₁individual electric current, by described k ₁individual electric current is input to k ₁in individual data input point, now by Q _μ[k ₁] being labeled as the input data that traveled through, μ ∈ { 1,2,3 ... A * n ₁; Store the output valve of obtaining on data output point simultaneously.

Steps A 6, judge above-mentioned A * n ₁whether individual input data are all the input data that traveled through, and if so, explanation is directed to each input data and has obtained an output valve, enters steps A 7, otherwise, return to steps A 5.

Steps A 7, judgement said n ₂whether individual structure data are all the structure data that traveled through, and if so, enter steps A 8, otherwise return to steps A 4.

Because need to be to n ₂individually for building the structure data in magnetic field, judge, to choose the optimum data that build, therefore need to build data for each, utilize selected A * n ₁individual input data are trained.

Steps A 8, obtain this circulation in n ₂the corresponding n of individual structure data ₂group output valve, wherein, every group has A * n ₁individual output valve, Yi Zuwei unit judges that each output valve, whether its corresponding input in the determined output area of data type, if so, makes B _jconstant, otherwise make B _jequal B _jadd difference DELTA, wherein said difference DELTA is: in the corresponding output area of output valve with the difference of the immediate value of output valve and output valve square.

Enumerating a detailed example below describes

If obtain 5 types of data in mode identification procedure, every kind of output area corresponding to input data is that the output area of the first input data is [0,20], the output area of the second input data is [20,40], the output area of the third input data is [40,60], the output area of the 4th kind of input data is [60,80], the output area of the 5th kind of input data is [80,100].For 20 of every type of data decimation, as input data, altogether comprise 100 input data, above-mentioned 100 corresponding output valves of input data are judged.If build data for first group, when first in input data is the first type, its output valve is 25.3, and it is no longer within [0,20] scope, so make B ₁equal B ₁add | 25.3-20|, continue second in input data to judge, the like.

The margin of error B that steps A 9, selecting step A8 obtain _jone of middle minimum, is designated as B _min, and judge B _minwhether be less than default ξ, if enter steps A 10; Otherwise, enter steps A 11.

Steps A 10, judge whether iterations iter is less than maximum iteration time Maxiter, if so, make iterations iter add 1, according to B _jutilize particle cluster algorithm to n ₂individual structure data O _j[k ₂] upgrade, enter steps A 3, if not, steps A 11 entered.

Wherein, ξ is the threshold value of setting as required; Because when minimum overlay amount sum is less than after ξ, explanation is now according to B _minthe corresponding constructed magnetic field of structure data can be realized input data is classified comparatively accurately, otherwise enter steps A 3, continues to find optimum structure data.Pg in the present invention in particle cluster algorithm used builds in data preferably one for this circulates adjacent two, and adjacent two build corresponding B in data _jless one, pb circulated j to build data and j of current circulation and build in data preferably one last time, i.e. corresponding B in two structure data _jless one.

Steps A 11, general be B now _mincorresponding structure data build data as optimum, utilize the optimum data that build to choose k as the density value of electric current ₂individual electric current, by described k ₂individual electric current is input to k ₂in the input point of individual structure magnetic field, successively by Q _i[k ₁] from k ₁in individual data input point, input, and according to the residing scope of output valve, the data of input are classified.

In sum, these are only preferred embodiment of the present invention, be not intended to limit protection scope of the present invention.Within the spirit and principles in the present invention all, any modification of doing, be equal to replacement, improvement etc., within all should being included in protection scope of the present invention.

Claims (2)

1. based on magnetic field, realize a method for simulation yard computer data classification, for Ψ the k that mode identification procedure is obtained ₁dimension data Q _i[k ₁] classify, wherein, i=1,2 ... Ψ, and Q _i[k ₁] in comprise A type data; It is characterized in that, concrete steps are:

Step 1, on magnetic conductive media film, choose M ₁* M ₂individual point vertically arranges probe; From set probe, choose k2 probe as building magnetic field input point, choose k ₁individual probe, as data input point, is chosen a probe as data output point; Wherein, building magnetic field input point, data input point and data output point is mutually different point;

Step 2, maximum iteration time is set is Maxiter, and sets the corresponding output area of each type in A type of data, with seasonal iterations iter=0;

Step 3, utilize particle cluster algorithm to generate at random n ₂individual for building the structure data O in magnetic field _j[k ₂], j=1,2 ... n ₂, each dimension that builds data is k ₂, n ₂individual structure data are different between two, and each builds the corresponding margin of error B of data _j, make B _j=0; From Q _i[k ₁] each categorical data in choose respectively n ₁individual data are as input data;

Step 4, from n ₂individual O _j[k ₂] in choose a k who is not traversed ₂dimension builds data, is designated as O _η[k ₂], by O _η[k ₂] middle k ₂each dimension value of dimension data is as the density value of electric current, by k ₂individual electric current is input to respectively k ₂in the input point of individual structure magnetic field, now by described O _η[k ₂] be labeled as the structure data that traveled through;

Step 5, from A * n ₁in individual input data, choose a k who is not traversed ₁dimension input data, are designated as Q _μ[k ₁], by Q _μ[k ₁] middle k ₁each dimension value of dimension data is as the density value of electric current, by k ₁individual electric current is input to respectively k ₁in individual data input point, now by described Q _μ[k ₁] be labeled as the input data that traveled through; Store the field strength values obtained on data output point or magnetic induction density value as output valve simultaneously;

Step 6, judge above-mentioned A * n ₁whether individual input data are all the input data that traveled through, and if so, enter step 7, otherwise, return to step 5;

Step 7, judgement said n ₂whether individual structure data are all the structure data that traveled through, and if so, enter step 8, otherwise, return to step 4;

Step 8, obtain this circulation in n ₂the corresponding n of individual structure data ₂group output valve, wherein every group has A * n ₁individual output valve, Yi Zuwei unit judges that each output valve is whether in the determined output area of its corresponding input data type, if so, B _jconstant, otherwise make B _jequal B _jadd difference △, wherein said difference △ is: in the corresponding output area of output valve and the difference of the immediate value of output valve and output valve square;

Step 9, choose margin of error B _jone of middle minimum, is designated as B _min, and judge B _minwhether be less than default ξ, if enter step 11; Otherwise, enter step 10;

Step 10, judge whether iterations iter is less than maximum iteration time Maxiter, if so, make iterations iter add 1, according to B _jutilize particle cluster algorithm to n ₂individual structure data O _j[k ₂] upgrade, enter step 4, if not, step 11 entered;

Step 11, general be B now _mincorresponding structure data build data as optimum, utilize the optimum data that build to choose k as the density value of electric current ₂individual electric current, by k ₂individual electric current is input to k ₂in the input point of individual structure magnetic field, successively by Q _i[k ₁] from k ₁in individual data input point, input, and according to the residing scope of output valve, the data of input are classified.

2. the method for Data classification according to claim 1, is characterized in that, described M ₁=M ₂, and the distance between every adjacent 2 is equal, M ₁* M ₂individual point has formed a rectangular array.

Images