CLASSIFICATION.EXAMPLE		3/10/89 JF	THIS IS A RESULTS FILE PRODUCED BY A RUN OF CA CLA WITH 50S DATA SET
												COMMENTS APPEAR HERE
												====================
SPIDER V5 (01/16/86)/(ALBANY VX750 ) ON 14-JUL-88 AT 16:42:0
SPIDER 5
*** PROJECT CODE: DE5   DATA CODE: FIA ***
    WELCOME    
%%%   TO    %%%
+++  THE    +++
...  WORLD  ...
      OF       
*** SPIDER  ***
.OPERATION: 
    B23    
.OPERATION:
    B23                                                                             
** START OF B23.DE5
                                           **
     1    CA CLA                                                                          
     2    7                                                                               
     3    CLU007                                                                          
     4    1-4                                                                             
     5    5,3                                                                             
     6    4                                                                               
     7    0.000000                                                                        
     8    2.000000                                                                        
     9    Y                                                                               
    10    DDG007                                                                          
    20    EN                                                                              
.OPERATION: 
    CA CLA    

.INP FILE CODE: 
          7      0										The file code identifying the
.CLUST  FILE: 											CORAN output files 
    CLU007											Cluster output file.
.FACTOR NUMBERS: 
    1-4 											Iterations = iterations of the
.# OF ITER./PART., # CENTERS/PART.: 								K-means algorithm.
          5      3										Centers = images chosen
. # OF PARTITIONS: 										randomly from which the K-means
          4      0										algorithm starts.
                                                    STEP    ** CLASSY **			
------------------------------------------------------------------------------------------------------------------------------
SPECIFICATIONS FOR   CLASSY
          FACTORS USED :
             1   2   3   4
          NBASE=   4     NITER=   5     NCLAS=   3     NKLA = 100
MEMORY RESERVATION          YOU HAVE RESERVED100000          YOU NEED  7752                         CLUSTERING BY AGGREGATION AROUND MOBILE CENTERS
                        PARTITION OF    100  OBJECT CHARACTERIZED BY    4  CARTESIAN COORDINATES
------------------------------------------------------------------------------------------------------------------------------
             PARTITION CONTAINS  100  CLASSES
          THE   99 FIRST CONTAINS THE MOST STABLE OBJECTS IN THE    4  BASIC PARTITIONS		Summary of parameters
          EACH PARTITION IS GENERATED BY    5  ITERATIONS AROUND    3  SEED-OBJECTS DRAWN AT RANDOM	specified
.ENTER SEED INTEGER (0=RANDOM DRAW): 
            0.000000
** RANDOM SEED ASSIGNED =       601393 								Use this integer when
													you wish to precisely
          CONSTRUCTION OF A PARTITION WITH SEED-OBJECTS    68    13    35				reproduce this result!
                    SIZE OF CLUSTERS AFTER   5 ITERATIONS                                               partition 1
                                                                          66.   17.   17.
          CONSTRUCTION OF A PARTITION WITH SEED-OBJECTS    27    24    10
                    SIZE OF CLUSTERS AFTER   5 ITERATIONS                                               partition 2
                                                                          33.   35.   32.
          CONSTRUCTION OF A PARTITION WITH SEED-OBJECTS    93    15    60
                    SIZE OF CLUSTERS AFTER   5 ITERATIONS                                               partition 3
                                                                          35.   33.   32.
          CONSTRUCTION OF A PARTITION WITH SEED-OBJECTS     1    20    48
                    SIZE OF CLUSTERS AFTER   5 ITERATIONS                                               partition 4
                                                                          35.   32.   33.
                                                                                                   
          SIZE OF THE    81 CLUSTERS FROM THE CROSSED-PARTITION                                    81=3**4 possible clusters
         FOLLOWED BY THECUMULATIVE PERCENTAGES.

    34    32    17    16     1                                                                     of these, only five contain
                                                                                                   any objects!  One cluster
  34.0  66.0  83.0  99.0 100.0                                                                     has a single member only
         SIZE OF RESIDUAL CLUSTER   (NUMBER 100 )=     0
                                              PERCENTENTAGE =     0.00
-------------------------------------------------------------------------------------------------------------------------------
  1   1   1   1   1   1   1   1   1   1   1   1   1   1   1   1   1   1   1   1                    cluster assignments of the
  1   1   1   1   1   1   1   1   1   1   1   1   1   1   3   3   4   4   3   3                    100 objects (20 per row,
  4   5   3   4   3   4   4   3   4   3   4   3   4   3   4   3   4   3   4   3                    from 1 to 100, numbers refer
  3   4   3   4   3   4   3   4   2   2   2   2   2   2   2   2   2   2   2   2                    to cluster number)
  2   2   2   2   2   2   2   2   2   2   2   2   2   2   2   2   2   2   2   2
.PERC. FOR CLASS CUTOFF(0=NO CUTOFF): 
            2.000000
DESCRIPTION OF THE HIERARCHY NODES                                                                  At this stage a decision
                                                                                                    is made on which clusters
                                                                                                    to include in the 
 NO SENIOR  JUNIOR  NO.  WEIGHT    INDEX                                                            hierarchical merging
                                                                                                    process. The cutoff (2%) 
  5     3     4     2    33.00   0.0049 ****                                                        specified excludes the 
  6     1     2     2    66.00   0.0331 *********************                                       single-member cluster
  7     5     6     4    99.00   0.1457 **************************************************************************************
DO YOU WANT DENDROGRAM PLOT FILE? (Y/N): 
    Y                                                                                     Result of hier. classification:
.ENTER FILE NAME FOR DENDROGRAM:                                                          Since only four clusters have passed
    DDG007                                                                                the cutoff, the merged clusters
File already exists, O.K. to erase and overwrite? (N/Y):                                  (classes) are assigned numbers 5 and
    Y                                                                                     up.
                                                                                          The description of the hierarchy 
DELETE DDG007.FIA;*                                                                       nodes contains the steps of merging
FILE OPENED: DDG007.FIA                                                                   each of the original clusters with
                                                                                          the group of clusters already merged.
                                                                                          In our example, the original clusters
                                                                                          3 and 4 are found to be closest, and
                                                                                          they are merged to form the new 
										   cluster #5. Similarily, 1 and 2 are 
										   merged to form #6. The last step 
										   (which is always trivial ) merges 
										   the new clusters #5 and #6 to form 
										   the trivial cluster #7, which 
										   contains all objects.
NODE  INDEX  SENIOR  JUNIOR  SIZE     DESCRIPTION OF THE HIERARCHY CLASSES                
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 
                                                                                          The hierarchy index gives the 
										   decrease 
   5    0.005    3     4      2         3    4                                            in intra-cluster variance resulting
                                                                                          from the merging. A small decrease
                                                                                          (as in 3+4) indicates that the 
   6    0.033    1     2      2         1    2                                            clusters are in close proximity, 
                                                                                          and can practically be regarded 
                                                                                          as the same cluster. A large 
   7    0.146    5     6      4         3    4    1    2                                  decrease (as in 1+2) indicates 
										   that the clusters are truly 
                                                                                          distinct, and should only be 
										   regarded as a single cluster 
										   relative to even more distant 
										   configurations.
     WEIGHT    INDEX         DENDROGRAM    (SCALE     0.00   0.15 )
    32.000     0.033    2  .......................                                                                                 
                                                 .                                                                                 
    34.000     0.146    1  ..................................................................................................
                                                                                                                            .
    16.000     0.005    4  ...                                                                                              .
                             .                                                                                              .
    17.000    ------    3  ..................................................................................................
                                                                                                                             
-----------------------------------------------------------------------------------------------------------------------------
                                                END OF STEP      ** CLASSY **
-----------------------------------------------------------------------------------------------------------------------------
LIST OF CLASS MEMBERS
CLASS
    1   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15  16  17  18  19  20         note that a single object, #42
       21  22  23  24  25  26  27  28  29  30  31  32  33  34                                 is missing, because its cluster
                                                                                              was rejected based on the 2%
    2  69  70  71  72  73  74  75  76  77  78  79  80  81  82  83  84  85  86  87  88         cutoff.
       89  90  91  92  93  94  95  96  97  98  99 100
    3  35  36  39  40  43  45  48  50  52  54  56  58  60  61  63  65  67
    4  37  38  41  44  46  47  49  51  53  55  57  59  62  64  66  68
LIST OF CLASS CENTER COORDINATES
CLASS  SIZE    1       2       3       4                                 This gives a quick idea of how many factorial axes
    1     34   -0.0120  0.0227 -0.0028  0.0002                           are involved in the distinctions. In this example,
    2     32   -0.0285 -0.0188  0.0003  0.0003                           factor 1-3 are needed to get a good picture of what
    3     17    0.0424 -0.0067 -0.0096 -0.0037                           is happening.
    4     16    0.0366 -0.0050  0.0131  0.0033
RE-CLASSIFICATION LOOKUP TABLE
ORIGINAL CLASS                                                           only two non-trivial cuts can be made in the dendo-
                                                                         gram to decide on class membership: cut #1 leads to
         1  2  3  4                                                      two classes (row #2 of table gives classification
    2    1  1  2  2                                                      of original clusters into 1 or 2); cut #2 leads to
    3    1  2  3  3                                                      3 classes (row #3 of table gives classification into
                                                                         1, 2, and 3).

DISPERSIONS AND INTER-CLASS DISTANCES OF 10 LARGEST CLUSTERS
CLASS      DISP         NEIGHBORS             1      2      3      4      5      6      7      8      9     10
          1    0.0191       2  4  3                                          For each cluster, the dispersion and the distance
          2    0.0204       1  4  3       0.0448                             to the other clusters is calculated. In addition,
          3    0.0178       4  1  2       0.0623 0.0727                      the three nearest neighbours (in the order of
          4    0.0128       3  1  2       0.0583 0.0678 0.0244               increasing distance) are printed out.
                                                                             Dispersion:
                                                                             D=1/N*sum(xi-xc)**2    (sum from 1 to N)
                                                                             (xi-xc)**2 is the squared euclidean distance
                                                                             between object i and the cluster center c.
.OPERATION: 
    EN    

 COMPLETED   14-JUL-88 AT 16:42:41