supraHex 1.13.3

How to control the layout when using visCompReorder?

Notes:
  • All results are based on supraHex (version 1.13.3).
  • R scripts (i.e. R expressions) plus necessary comments are highlighted in light-cyan background, and the rest are outputs in the screen.
  • Images displayed below may be distorted, but should be normal in your screen.
  • Functions contained in supraHex 1.13.3 are hyperlinked in-place and also listed on the right side.
  • Key texts are underlined, in bold and in pumpkin-orange color. 
    •       
    # The layout shown in visCompReorder cannot be controlled by itself. The same thing is true for the side legend, which is automatically changed to fit the overall layout. 

# Instead, it is controlled by the sister function sCompReorder, from which xdim (and/or ydim) can be specified to layout the dimension of the rectangular lattice.

# generate an iid normal random matrix of 100x10
data <- matrix( rnorm(100*10,mean=0,sd=1), nrow=100, ncol=10)
colnames(data) <- paste('S', seq(1:10), sep="")
# get trained using by default setup
sMap <- sPipeline(data=data)

    
Start at 2017-03-27 18:59:33

First, define topology of a map grid (2017-03-27 18:59:33)...
Second, initialise the codebook matrix (61 X 10) using 'linear' initialisation, given a topology and input data (2017-03-27 18:59:33)...
Third, get training at the rough stage (2017-03-27 18:59:33)...
	1 out of 7 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	2 out of 7 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	3 out of 7 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	4 out of 7 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	5 out of 7 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	6 out of 7 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	7 out of 7 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
Fourth, get training at the finetune stage (2017-03-27 18:59:33)...
	1 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	2 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	3 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	4 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	5 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	6 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	7 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	8 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	9 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	10 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	11 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	12 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	13 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	14 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	15 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	16 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	17 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	18 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	19 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	20 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	21 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	22 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	23 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	24 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
	25 out of 25 (2017-03-27 18:59:33)
	updated (2017-03-27 18:59:33)
Next, identify the best-matching hexagon/rectangle for the input data (2017-03-27 18:59:33)...
Finally, append the response data (hits and mqe) into the sMap object (2017-03-27 18:59:33)...

Below are the summaries of the training results:
   dimension of input data: 100x10
   xy-dimension of map grid: xdim=9, ydim=9, r=5
   grid lattice: hexa
   grid shape: suprahex
   dimension of grid coord: 61x2
   initialisation method: linear
   dimension of codebook matrix: 61x10
   mean quantization error: 4.9630432442287

Below are the details of trainology:
   training algorithm: batch
   alpha type: invert
   training neighborhood kernel: gaussian
   trainlength (x input data length): 7 at rough stage; 25 at finetune stage
   radius (at rough stage): from 3 to 1
   radius (at finetune stage): from 1 to 1

End at 2017-03-27 18:59:33
Runtime in total is: 0 secs


    # reorder component planes
sReorder <- sCompReorder(sMap=sMap)

    
Start at 2017-03-27 18:59:33

First, define topology of a map grid (2017-03-27 18:59:33)...
Second, initialise the codebook matrix (30 X 61) using 'linear' initialisation, given a topology and input data (2017-03-27 18:59:33)...
Third, get training at the rough stage (2017-03-27 18:59:33)...
	1 out of 300 (2017-03-27 18:59:33)
	30 out of 300 (2017-03-27 18:59:33)
	60 out of 300 (2017-03-27 18:59:33)
	90 out of 300 (2017-03-27 18:59:33)
	120 out of 300 (2017-03-27 18:59:33)
	150 out of 300 (2017-03-27 18:59:33)
	180 out of 300 (2017-03-27 18:59:33)
	210 out of 300 (2017-03-27 18:59:33)
	240 out of 300 (2017-03-27 18:59:33)
	270 out of 300 (2017-03-27 18:59:33)
	300 out of 300 (2017-03-27 18:59:33)
Fourth, get training at the finetune stage (2017-03-27 18:59:33)...
	1 out of 1200 (2017-03-27 18:59:33)
	120 out of 1200 (2017-03-27 18:59:33)
	240 out of 1200 (2017-03-27 18:59:33)
	360 out of 1200 (2017-03-27 18:59:33)
	480 out of 1200 (2017-03-27 18:59:33)
	600 out of 1200 (2017-03-27 18:59:33)
	720 out of 1200 (2017-03-27 18:59:33)
	840 out of 1200 (2017-03-27 18:59:33)
	960 out of 1200 (2017-03-27 18:59:33)
	1080 out of 1200 (2017-03-27 18:59:33)
	1200 out of 1200 (2017-03-27 18:59:33)
Next, identify the best-matching hexagon/rectangle for the input data (2017-03-27 18:59:33)...
Finally, append the response data (hits and mqe) into the sMap object (2017-03-27 18:59:33)...

Below are the summaries of the training results:
   dimension of input data: 10x61
   xy-dimension of map grid: xdim=6, ydim=5, r=3
   grid lattice: rect
   grid shape: sheet
   dimension of grid coord: 30x2
   initialisation method: linear
   dimension of codebook matrix: 30x61
   mean quantization error: 4.58202954421453

Below are the details of trainology:
   training algorithm: sequential
   alpha type: invert
   training neighborhood kernel: gaussian
   trainlength (x input data length): 30 at rough stage; 120 at finetune stage
   radius (at rough stage): from 1 to 1
   radius (at finetune stage): from 1 to 1

End at 2017-03-27 18:59:33
Runtime in total is: 0 secs


    # visualise multiple component planes reorded within a sheet-shape rectangle grid
visCompReorder(sMap=sMap, sReorder=sReorder)

# As you have seen, these 10 samples are organised onto a 6 X 5 rectangle grid. You can change it to 7 X 7, for example:
sReorder <- sCompReorder(sMap=sMap, xdim=7, ydim=7)

    
Start at 2017-03-27 18:59:34

First, define topology of a map grid (2017-03-27 18:59:34)...
Second, initialise the codebook matrix (49 X 61) using 'linear' initialisation, given a topology and input data (2017-03-27 18:59:34)...
Third, get training at the rough stage (2017-03-27 18:59:34)...
	1 out of 490 (2017-03-27 18:59:34)
	49 out of 490 (2017-03-27 18:59:34)
	98 out of 490 (2017-03-27 18:59:34)
	147 out of 490 (2017-03-27 18:59:34)
	196 out of 490 (2017-03-27 18:59:34)
	245 out of 490 (2017-03-27 18:59:34)
	294 out of 490 (2017-03-27 18:59:34)
	343 out of 490 (2017-03-27 18:59:34)
	392 out of 490 (2017-03-27 18:59:34)
	441 out of 490 (2017-03-27 18:59:34)
	490 out of 490 (2017-03-27 18:59:34)
Fourth, get training at the finetune stage (2017-03-27 18:59:34)...
	1 out of 1960 (2017-03-27 18:59:34)
	196 out of 1960 (2017-03-27 18:59:34)
	392 out of 1960 (2017-03-27 18:59:34)
	588 out of 1960 (2017-03-27 18:59:34)
	784 out of 1960 (2017-03-27 18:59:34)
	980 out of 1960 (2017-03-27 18:59:34)
	1176 out of 1960 (2017-03-27 18:59:34)
	1372 out of 1960 (2017-03-27 18:59:34)
	1568 out of 1960 (2017-03-27 18:59:34)
	1764 out of 1960 (2017-03-27 18:59:35)
	1960 out of 1960 (2017-03-27 18:59:35)
Next, identify the best-matching hexagon/rectangle for the input data (2017-03-27 18:59:35)...
Finally, append the response data (hits and mqe) into the sMap object (2017-03-27 18:59:35)...

Below are the summaries of the training results:
   dimension of input data: 10x61
   xy-dimension of map grid: xdim=7, ydim=7, r=4
   grid lattice: rect
   grid shape: sheet
   dimension of grid coord: 49x2
   initialisation method: linear
   dimension of codebook matrix: 49x61
   mean quantization error: 3.08127685132509

Below are the details of trainology:
   training algorithm: sequential
   alpha type: invert
   training neighborhood kernel: gaussian
   trainlength (x input data length): 49 at rough stage; 196 at finetune stage
   radius (at rough stage): from 1 to 1
   radius (at finetune stage): from 1 to 1

End at 2017-03-27 18:59:35
Runtime in total is: 1 secs



    visCompReorder(sMap=sMap, sReorder=sReorder)

# In the function sComReorder, rather than specifying dimensions it is better to change another parameter 'amplifier', which means amplifying the number of samples as a total of nodes in the rectangle grid. The rest work will be done by the function. Internally, once the total nodes are fixed, the xdim/ydim ratio is the square root of the two biggest eigenvalues of the input data.

    Source faq

    FAQ2.r

    Functions used in this FAQ