@rudzinskimaciej In biological neural networks there is this phenomenon of mapping dimensionality reduction & expansion.
For example in the cockroach there are 300K Kenyon cells feeding from a few hundred projection neurons.
For example in the cockroach there are 300K Kenyon cells feeding from a few hundred projection neurons.
@rudzinskimaciej The need for projecting into such a very high dimensional sparse space is to have simple linear readout neurons at the output & follows Cover's theorem.
@rudzinskimaciej *Cockroach olfactory network to be precise.
Loading suggestions...