Perhaps the limits of Eureqa's power?

[Damir Olejar]

I have tried some simple values, but got some disappointing results.

Input fields were 440Hz and 443Hz sine waves.
The output fields are representing a 3Hz sine wave.

Maximum and minimum amplitudes for all waves were a constant.

However, Eureqa cannot find the function for a 3Hz sin wave.


I have also tried going from 440hz square and saw-tooth wave to a
440hz sin wave, but no good results were obtained either.

Am I doing something wrong, or have I reached the limits for Eureqa ?




Thank you for an answer.

[RRogers]

If you save the project and post a link I will try it.
Ray

[Michael Schmidt]

I tried entering these two target expressions, just using the default dataset that loads when starting Eureqa/Formulize:

"sin(3*w) = f(w)"

"sin(440*w) = f(w)"

It finds both in under 60 seconds. Can you confirm?

Michael

--
Eureqa Formulize ( http://www.nutonian.com )
-------------------------------------------------
Unsubscribe: eureqa-group...@googlegroups.com
View Group: http://groups.google.com/group/eureqa-group

[Damir Olejar]

Without constants, the result for 443Hz-440Hz = 3Hz wave, gave

"R^2 Goodness of Fit" 0.99796389

"Correlation Coefficient" 0.99710621

This is alright and I don't know why it didn't work that well last night (there is a high-frequency overtone, but that can be cut-off with any sound-editing software).

Originally, I am trying to find a general formula for a waves that shift from (lets say) 20Hz to 21000Hz and 23Hz to 21003Hz within a same time-frame, to give a resulting wave of 3Hz, without much success except for pretty patterns on the Residual Error graph.  

Idea is to get the formula for binaural frequencies from two sound-waves (or to insert a binaural frequencies into waves).

Furthermore, I am trying to find a formula that would flip the fourier analysis of a sound into a negative, and also a formula that would subtract the sounds by frequency cut-off rather than by subtracting amplitudes... by using the bitmap and waves program, with a "darken" by layers in photoshop (then converting it back to sound to get the formula).

http://www.elisanet.fi/victorx/BitmapPlayer.htm

I will post the data files as TXT, and some screen-shots as explanations, as soon as I get some free time (tomorrow I guess, since I got a lot of junk in the folder).

--
Blue butterfly ants, chanted romantically.

[Damir Olejar]

Here is a data-set I am trying to find a formula with but without a success (attached: eq.txt.rar  It is a TXT file to be imported by EQ).

Going from left to right....

First column:

Input wave 1

Second column:

Input wave 2

Third column:

Output

Input wave 1 and 2 are the sound-waves and their amplitudes. The formula should choose the "quieter" frequency from two inputs.

The problem could be that data-set is too large to work with.  When I reach the promising result with a much smaller portion of that data-set, it loses its value when I introduce the whole data-set back again. This goes on and on, and it could be possible that EQ is not a good program for finding this formula, or that I need to approach the problem differently.

Other problem could be that the program that I used to find the waves applies the wavelets (http://www.sonicspot.com/coagula/coagula.html). However, I have made sure that all data is processed with the same program to avoid the data mismatch.

The other data I mentioned (for finding the binaural frequencies) is something I will investigate further on my own.

Thank you kindly,

Damir

[Dave Nunez]

Hi there, I don't really understand what you are trying to achieve but
I took a look at your data and it seems that it needs some
preprocessing before giving it to EQ. I tried feeding the raw data to
EQ, and it found a lot of outliers. So try to preprocess it first.
can you please clarify what you mean when you say"quieter".

Beers, -d

> --
> Eureqa Formulize ( http://www.nutonian.com )
> -------------------------------------------------
> Unsubscribe: eureqa-group...@googlegroups.com
> View Group: http://groups.google.com/group/eureqa-group

--
When things get too complicated, it sometimes makes sense to stop and
wonder: Have I asked the right question?
-Enrico Bombieri

[Damir Olejar]

By quieter I mean the following:

If you have 2 frequencies, 

1Khz at 130 decibels 

3Khz at 40 decibels

Both appear at the same time by different input-sources. The quieter frequency would be 3Khz.

I am not sure about the outliers, since programs gave very precise digital outputs that I wouldn't like to change. 

I still believe fixing it would not improve the formula-finding.

[aidabet]

Hello Damir,
You are correct, when the dataset is too large, Eureqa just can't find that magic formula we want.
When you write

"and it could be possible that EQ is not a good program for finding this formula"

It's quite likely there is no close fitting formula.
Eureqa is still producing the best fit possible on that data, but it's
not the solution we hope to see, sometimes the closest fitting column
in the dataset is the best fit.
Nevi

DO> Here is a data-set I am trying to find a formula with but without a
DO> success (attached: eq.txt.rar It is a TXT file to be imported by EQ).

DO> Going from left to right....
DO> First column:
DO> Input wave 1

DO> Second column:
DO> Input wave 2

DO> Third column:
DO> Output

DO> Input wave 1 and 2 are the sound-waves and their amplitudes. The formula
DO> should choose the "quieter" frequency from two inputs.

DO> The problem could be that data-set is too large to work with. When I reach
DO> the promising result with a much smaller portion of that data-set, it loses
DO> its value when I introduce the whole data-set back again. This goes on and
DO> on, and it could be possible that EQ is not a good program for finding this
DO> formula, or that I need to approach the problem differently.

DO> Other problem could be that the program that I used to find the waves
DO> applies the wavelets (http://www.sonicspot.com/coagula/coagula.html).
DO> However, I have made sure that all data is processed with the same program
DO> to avoid the data mismatch.

DO> The other data I mentioned (for finding the binaural frequencies) is
DO> something I will investigate further on my own.

DO> Thank you kindly,

DO> Damir