Sub Label in a Labelset

Hello Flucoma!

I’m working on a automatic folder classifier, I’m training a neural network to be able to place samples in different folders according to the labels used in it. Right now I’m focusing on the training part of the device and I’m having a problem. With this classifier, I wish to be able to have different labels and sublabels. Is it possible in a labelset to link different label together so that i can hierarchize them and recall them with a funneling method. For example, I have a master label call Harm and i would like to have 3 sublabel called piano, guitar and synth. My goal with these sublabels and labels would be to be able to search more or less precisely in the folders.

Thanks in advance for any recommandation !
RJ

here is the code i made for the neural network training


----------begin_max5_patcher----------
2816.3oc2cs0aaajE9YGf9efPnuspzy8K8ozcCJ1EnODrc2mJJLnjFovrRjB
jT4BJZ9sumYnjrsHE0P6YjahSfULE0Ly4aNW9Nm4Pm+36d0MSlU9IS8jjeL4
2Rt4l+.txMtqYuxMGtvMS1j8o4qypc23jMl55rUlIS2+lMlO03disU4EMGub
wtM4EqMMtOD4AWsbWygKiOb41q074sl10xjII+9g2KegazKm89efPON7ayZl
+t7hU2UYl2z9gDLcJAIHbtPxnBBVQmlHkxTrRyIXgDQYXllLMgiRQSSHjTTx
u6Fu+76dk8U3kodiCElOBqotvv5xrEyxJV0ORfGKR3Fp9QC9PnAWfSYTkhnY
DrVKQxoILsJEKoLoDvJgRII3oIXz2DnAEODZnooLIUnvZhPRjDM2hFjTLRKX
J.hzvaqTgBMNqMRkIaQx+LqZS4cu0TM+MSSVray1i2257By7xcEtalFSSIJa
rvEmzmxCFagKqAULfqOVk2XhpKE5fFQTdO3.izEGnrqnVyu7xo0HFqKmuQ0Z
jChCDTOpMbQjTaNmq2Zy5D7nfAxv9cmdN3fMXbYphlxUTNWhIJsM.MfFJd2v
xJTpVvwLICdCgVJ4Oaroob0p0lKD7YaVU1FSio5NSQ1r0NYFMVEEK2m9AG0P
fCgnSYHpVJzXIQHTLADYBLjTXgfAv.yRngoAjnUWg8rviU0eLGVC8CHGccj2
5JIr.zYPG8nQGNnSzQ2gHnoHPmAwQHMSfXVHCHDpUTMmqkZvxCFsHXlMubyL
vAbxVvybZRdxqapxWsxTUmfBjw24M73nA8CwXco.B9i6nZgkwzOTUBHnlpxe
May1yYKFHsI9fj.EZcG7fpEoLPmAxX.yQRLFIhNIvUll1v0wLDEePBdLhLkf
4BvuLBvAN2Jz83Rlyinhw1x0ed1tkKMUewY8LpbDF1b474IvG1gipGFeXdOH
CnobOzbXDqy9fYwcvz.C3cYMfmfY6ZZSv9l6QmalX1Ly3VRn1qrGUCJ51jLK
oNPP5vg+ECxJjRRYNWMDsBbjqz1rrfDL5hnwO0qkkqWXpR9G+3s+2ZvC8sUl
Me9829lx4fCphl5a+O.ww0Yf27a+406.G6Y28Vq.c6OUjs9y0l+tSYEVe29u
+by61zOAbQLspEpQq8ZYYiDfKeHDnFfali.N0B0ZbbfZWpJQ06lXPqXFimxY
JPVEDFVJPVJmBrNRUB5r3vaOqSs.ACxQS7Ndnv47C89x7h3BBWHE9tIkJr4h
0gAoNkGMP3GFYlXOg7LjC5EFr308DYSBfCgRTRsRJQZDyBAhHpMbM.B4v.AV
1CPvv8kSQDAh+1U.HFzIIFn80EHH3qMPb8JSghOdECRerhY33ltzLfnQTySR
IeB9J.BucQBcDQBGkJSU+HAuejfMnYxz8ISOc.qFkZ7VM85GUDBsj+BT.K0S
vMRejNCRArNmtBH9Uk1THBjS0ym.oFMd3PgSoRPqfRYHAPEwdxjw8b2pGxKx
Yx+C0mzRtfmhtBKEBgnjT3GTDNBCBqHxNJ1UrMa9+KoNIOfkP3r1CZ5kR2sS
PDoHkxjfOSpjADRATI5o65bdlfhJyaMezHAgesSB6pTsM8vjuknNUaqunoAo
ZayK2XKlQWf3MYMY+poIA96ujMyr19uWrK4m9WOemDCSm.nUgvTHKTFC3XQs
hOUSR4TNQI.+gZrf5NVv8JBn3H+YsEwIYYNrBOGshwH0pKH0bjTxPbkPqEHo
BjZgU6GfA.LHH3EIDGPwCfPeVs+rEK1BIf2j783jZvLXwOaxZ1UYpewNC4GF
itObiwO0VQ.JKcXWgw7.bFxCU7T3OiRGgdwzRdzqmAbHisXFT6gHqnXpsMVj
XE2dBFBLCruXJNDAFbBiHDdDoo2d.Xtp9csOALLZzU+IhGA1Plgqs9bslgeO
YHaORbs8DitVY8Y7ocfEU7094EhQC5CGLk5ZtIIw5.CGfK+EAjQD2BitPBcZ
UJPLggnVaDkRSbgtb0R9wdYhME10lrpSrUhhUAdvHRLgnaek1qYgD+sPmThw
pQSqW808w40ZYU+BbfdXxnyghq66zni5A5ALGqlaffWw0NjLXzIMSjp6.FPb
aBmR4HJXFpHJKQGIMD1g61L6hYILhJu8TaXAL4ITPRJhekOluCr.cIYT+XZf
33SCjLXLMcucjYuslZP3AtHedS5GxMe74GrlhFKiMkBDAf1OjgDHuZJFQOJW
X0ecDL7HSCre4RHBgb8xyAgNXheBNv9BgDbpPKPBDV8RPAoBDFy3whmjAMkN
Z8dFpG8inxLe45c4KRcY1UaZ9RRiot4oUB54kqKqZkMTJ0VmLASh.12DpsXG
VY.kRjPlGZEhpslFZ20uG8FE5xFK5Js0opK5p9VQciOVuQuTZaKxZxbJaKJK
JLl8bYCfVGQpfbZg7a4LMhX02XfHqDGz5X1KIckx7388bTAEiDw6WCTDydS3
P2o9nJml75ZHcAfUCMPm1zRH0tlKPBb3m3DHLfusspVF.7Z1tllxhqwoRODj
LHuXgPae59.kWkfg0BoTZySU1sJFQ8boeXON2RM9pzkyXpdroU90YWN25SDP
3MKmO+KfMIXc1jU0LuzrbYBN40.dtzpGVacNCek75kGriO0p194b2a6sNlyF
AGvPQLzneJmEP9MJvqMr6YSFlAagVkcV.NgjsYEl0OM19yJqVXpdTLFGMlCe
6QgNdDBLXpAPT.P9zP5ZLICXIYS20JjcpNo846kXoLSQZIViTBWmX.asPvLN
ixzbrTc+Rnc89nM4MkKLORTAKtVOYDoUF1e0UUYKvGYu4HsAbmZIwYEV8C4s
YWFnTXa5gecBZXGPxgAj8fOJwJtvl8w.0fqMjxMf6+g8QmOY.2VUtsrpIG7Z
a2hR4mnPzdmNU8S+8AfaKv9FmnlzVTn85yGND7jGLmK.1v4EYGlze6AtlRFT
kz+Yj6yLZaXiPMg1gBewIjDtID4iDZ6Ys.MgJsWPJJbSnxKI7jc5ilpnq1RP
pOyR.Ghkfzqkf7keIHh4FgW1ymphtIusGGpOzTZjdJTLg12YTXoCSDTf6ilB
gHXJBaT2Zu2IBB71Y96dmwZaj6sGCRj1Fkd4CwUX6.4DQJ8VnYgYBEd60BGl
IzqsU2A7GJLk5k+fvIhB80dB8xuOOb7UDdo0bfUSnlPrWSXXfTtW6ghvokZe
Nmup6g1mwbOlPT3fTun.Z+Eof0gd.lPl+1gwhHfeKAWqDFM9PLk239K7RfFy
cB+hBDtTrXdkhkHbFzToeosxFj0IjveWRehd+sOGSzs5a9emRYOcNNUfR0md
hMcuwvfWhffWRuVwWYQyORVmh.mHZrdn+2qng4rT4IGxQrjLVHjLLh3oRt8A
pk8zuyWb3xqf7G.0.LgDuz7ngi2Dg9jTHBZ3MeWCrHFe6X4HuZLVcGbkOjaH
AbJ8JO.LkGvozSeo3.NkL+jRZ.mROUgQAbJIixxMDSIwyB5DxozOijSskN5a
fbEWDN3AGImj9tHP5X5kj3GoOV.0yI7q8wI457cOpOSLKAtmqgNnSX04vnQE
YNDSIxS07vUqpiEK8ZJkbuKpPnDRh20qJZ5zt0.9p58vMkjqZHK+P5Np.g0x
EgdNrC12F.yJ+zppxca64+Z.Nzd.22oF2+Lfe74htcbue7x1t8Clp58qg1Qa
xlr221XHpos+bdQ4C59hIUlOje3iz1HMSxpl+t7FybaG63ZWjOI12HqtN0np
XW9Afzhe1o10IKE1F.ZaVKR453ku6UvM7+83eq3I
-----------end_max5_patcher-----------

Hello @R_J and welcome,

I don’t completely follow what you want to here. Do you want to want to classify a bunch of sounds and use the classification to move / copy the files to folders? Or to look up matches within a hierarchy of folders based on some taxonomy?

In any case, I suspect that the answer could be to use a hierarchy of classifiers, e.g. a top-level that will distinguish between harm / perc, then something to classify classes within each of these, etc.

Hello @weefuzzy thank you for the quick response !

What i’m trying to do, like you said, is to look up matches within a hierarchy of folders. I will investigate what you’re suggesting me to use multiple fluid.mlpclassifier~ and to place them in a hierarchy.

Again thanks for your time and if I succeed i’ll post my code in this tread !