Year-end appeal: Please make a donation to the OEIS Foundation to support ongoing development and maintenance of the OEIS. We are now in our 61st year, we have over 378,000 sequences, and we’ve reached 11,000 citations (which often say “discovered thanks to the OEIS”).
%I #6 Mar 17 2018 04:12:04
%S 0,0,60,15652352,18446731528483929840,
%T 1461501637330902918203677267647731623106580665344,
%U 3940200619639447921227904010014361380507973
%N Imperfect formal neural networks with n components.
%C Number of {0,1}^n to {0,1}^n vector-vector maps of which at least one component is not a formal neuron, i.e., some are not threshold gates.
%D Labos E. (1996): Long Cycles and Special Categories of Formal Neuronal Networks. Acta Biologica Hungarica, 47: 261-272.
%D Labos E. and Sette M.(1995): Long Cycle Generation by McCulloch-Pitts Networks(MCP-Nets) with Dense and Sparse Weight Matrices. Proc. of BPTM, McCulloch Memorial Conference [eds:Moreno-Diaz R. and Mira-Mira J., pp. 350-359.], MIT Press, Cambridge,MA,USA.
%D McCulloch WS and Pitts W (1943): A Logical Calculus Immanent in Nervous Activity. Bull.Math.Biophys. 5:115-133.
%F a(n)=A057156(n)-A000609(n)^n=A057156(n)-A065246(n).
%e For n = 2 the 14 threshold gates determine 14*14 = 196 neural nets each built purely from threshold gates; the remaining 2^(2*4)-14^2 = 256-196 = 60 = a(2) functions are synthesized from both neurons and non-neurons. For n = 3, 104 = A000609(3) formal neurons and 152 non-neurons gives (2^24)-A065246(3) = 15652352 = a(4) nets with at least one linearly non-separable component.
%Y Cf. A000609, A065246, A065248, A064436.
%K nonn
%O 0,3
%A _Labos Elemer_, Oct 26 2001