Recent Posts

Pages: [1] 2
1
Symbolic Regression General / Re: Forecast of a new curve
« Last post by admin on June 03, 2021, 02:34:41 AM »
The best way would be to use an external tool to calculate the points and create this new plot -- gnuplot, Python+matplotlib, Excel, etc.

Creating new plots inside the program would be interesting, but this capability is not implemented at the moment.
2
Symbolic Regression General / Forecast of a new curve
« Last post by Giovanni Di Maria on May 30, 2021, 06:04:44 AM »
Hi
My goal is to forecast a new curve,not a single value.
I have 5 curves, respectively related with R parameter at R=0, 100, 200, 300, 400.
Is there a mode to create the curve number 6 (green color) with the new parameter R=500 ?
I am finding for the correct approach.
Thank you very much
Giovanni Di Maria
http://www.elektrosoft.it/hd/forecast.png
3
Vowwwwwww
Fantastic.
Thank you very MUCH!
Giovanni Di Maria
4
Symbolic Regression General / Re: Simple function but not interpolating
« Last post by admin on May 21, 2021, 12:32:31 PM »
To insert an attached image, you have to upload it somewhere else and insert the link between [ img][ /img] tags (remove the spaces)
5
Symbolic Regression General / Re: Simple function but not interpolating
« Last post by admin on May 21, 2021, 12:31:16 PM »
There are some things that can help speed the convergence:

1- Normalizing the data (dividing each column by its maximum value)
2- If you have some previous knowledge on which base functions are most relevant, only leave those enabled

If I do those two things and run the optimization with sin, cos, abs, +, *, and / as base functions, I quickly get this solution:

abs(0.922406*sin((2.37127*x*x+7.90349)*(-0.00619216+x))+0.0182209)*(1.2771-x)


The Hybrid (CC+RMS) search metric turned out to a good one for this case.
6
Now it seems OK.
I waited more time, about 3 hours.... for now a very good formula is:

(727.616+(-0.602032*x))*(0.0263795+cos(asin(cos((-0.0106622)*(104.588*tanh(fmod(0.00316671*x,-2.88173))-x)))-(0.000131321*(-183.887+x))))

Cheers
Giovanni Di Maria
7
Symbolic Regression General / Simple function but not interpolating
« Last post by Giovanni Di Maria on May 21, 2021, 08:30:46 AM »
Hello.
I have this dataset describing some kind of sine wave
(by the way, how do you insert an attached image in the forum?).

Why is TuringBot unable to interpolate this data well? The data trend is very simple, I could also do it manually.
Thank you

Giovanni Di Maria



x y
0 0
10 36.5
20 92.5
30 157.5
40 212.5
50 262.5
60 310.5
70 354.5
80 391.5
90 426.833
100 464.167
110 496.833
120 522.5
130 548.167
140 567.833
150 585.167
160 597.5
170 611.167
180 615.167
190 617.167
200 615.5
210 610.833
220 601
230 588.833
240 574.167
250 554.667
260 530.833
270 501.667
280 471.5
290 439
300 402.5
310 360.667
320 318
330 273.333
340 224.167
350 169.833
360 111.333
370 52.167
380 1.33299999999997
390 46.833
400 94.833
410 135.833
420 174.167
430 211.5
440 243.167
450 273.667
460 298.667
470 323.667
480 343.333
490 362.167
500 374.667
510 384.5
520 391.667
530 397.167
540 396.833
550 394
560 385.5
570 374.167
580 359.5
590 340
600 319.5
610 293.5
620 266.167
630 237.5
640 208.333
650 169.833
660 127.667
670 82
680 41.5
690 0.66700000000003
700 38.833
710 80
720 113.167
730 144.667
740 169.167
750 193.167
760 212
770 227.167
780 237.333
790 244.833
800 249.833
810 252.833
820 251.667
830 247.667
840 240.167
850 231.333
860 217.667
870 200.167
880 178.833
890 155.667
900 125.833
910 97.833
920 60.833
930 18.667
940 14.5
950 43.667
960 70.833
970 95
980 116
990 131.5
8
Symbolic Regression General / Re: Size of input numbers
« Last post by Giovanni Di Maria on May 20, 2021, 07:05:17 AM »
vowwwww
It's fantastic!!!!
Giovanni Di Maria
9
Symbolic Regression General / Re: Size of input numbers
« Last post by admin on May 19, 2021, 10:34:23 PM »
This is a great question. While the input file is parsed, the numbers encountered are converted to the "double" data type, which on most systems has a maximum value of around 10^308.

So as long as your numbers are not as big as

10000000000000000000000000000000000000000000000000000000000
00000000000000000000000000000000000000000000000000000000000
00000000000000000000000000000000000000000000000000000000000
00000000000000000000000000000000000000000000000000000000000
00000000000000000000000000000000000000000000000000000000000
00000000000000

you should be fine.
10
Symbolic Regression General / Size of input numbers
« Last post by Giovanni Di Maria on May 19, 2021, 05:23:48 PM »
Hi
What is the maximum size of input numbers?
Can i create, for example, this input file?

x y
1 700000000000
2 1400000000000
3 2100000000000
4 2800000000000

Thank you
Giovanni Di Maria
Pages: [1] 2