In my previous mail, I said:
> Basically, each output of your complex flowchart should be based
on at
> most one 'tmp' variable, and this 'tmp' variable should
correspond to a
> variable that, once evaluated, will have its chunks recombined
entirely.
Actually, I have just reworked the internal architecture of the
toolbox, and in my current version of this experimental attempt,
'tmp' variable can be computed in each chunk separately.
Does that mean I can only store one tmp variable at a time?
No, you can have several tmp variables.
Unfortunately my flow chart is indeed more complex, see Code below.
Note that quite a few features depend on the spectrogram which I
therefore would like to store in the tmp field. Can this flow chart
also be "reduced" to take advantage of the internal memory
management?
Maybe I can do something along the lines of
myflow.tmp.mspe = mirspectrum('Design', 'Frame', 512, 'sp', 1);
[calc all mspe dependent features]
[start a new flowchart and calculate the rest]
so that I at least do not have the spectrum calculated again and
again?
Yes, absolutely, you can do that. In the current version released
(1.2.2), the 'tmp' variable myflow.tmp.mspe would have been computed
across all chunks and stored in memory for the computation of the
dependent features. In the new version I am preparing, such 'tmp'
variable is computed in each chunk separately, and once the
dependent features are computed, it is erased from memory.
I will send you the new version once I have finished cleaning and
checking the code. Anyone else interested to receive this beta-
version can send me an email (olartillot@xxxxxxxxx).
Regards,
Olivier
Olivier Lartillot
Academy of Finland Research Fellow
Music Mining Plant
Finnish Centre of Excellence in Interdisciplinary Music Research
Department of Music
PO Box 35(M)
40014 University of Jyväskylä
Finland
http://users.jyu.fi/~lartillo/
Tel +358 14 260 1346
