On Friday, September 03, 2021, at 8:04 AM, magnuswootton81 wrote:
> a) Run it per sample, to get all overlaps.
Per each 44.1 kHz sample? Surely not? Also, do you do FFT first?
> b) do a knn match to a database,  if it isnt near enough to any of them,  you 
> store it as a new class.  (so its unsupervised learning, at least this part 
> of it.)
Hmm ok. I get this part, but I'm not sure what exactly the object is you put 
into the database. KNN is an interesting choice for a "dumb smart" algorithm 
(super simple but probably useful). Isn't it getting slow though over time?

I always like to hear about surprisingly simple algorithms.
------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T89dc799e249e2d0b-Mdd6055a601b53e564bf7b21a
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to