Computational Intelligence and Neuroscience

Research Article

Lateral Information Processing by Spiking Neurons: A Theoretical Model of the Neural Correlate of Consciousness

Algorithm 1

Algorithm which updates the state variables of neuron 𝑖 from one time step to the next.
(1) 𝑜 𝑖 ∶ = ( 1 − 𝛼 𝑜 ) 𝑜 𝑖 // output decay
(2) 𝑎 𝑖 ∶ = ( 1 − 𝛼 𝑎 ) 𝑎 𝑖 // activation decay
(3) 𝑎 𝑖 ∶ = 𝑎 𝑖 + 𝛼 𝑎 ∑ 𝑗 𝑤 𝑖 𝑗 𝑜 𝑗 // sum over all inputs
(4) // average activation across open gap
(5) // junctions (reconfigurable resistive grid)
(6) 𝑎 𝑖 ∶ = a v g ( { 𝑎 𝑖 } ∪ { 𝑎 𝑗 ∣ N e u r o n 𝑗 is connected to
(7) non-refracting neuron 𝑖 via open gap junction;
(8) 𝑎 𝑗 is then set to average 𝑎 𝑖 } )
(9) 𝑎 𝑖 ∶ = m a x [ − 1 , 𝑎 𝑖 ] // limit activation
(10) // reduce threshold based on size of sub-network
(11) 𝑡 𝑖 ∶ = m a x [ 0 , 1 − 𝛾 ⋅ 𝑁 𝑠 ]
(12) i f ( 𝑎 𝑖 > 𝑡 𝑖 ) { // neuron fires if above firing threshold
(13) 𝑎 𝑖 = 0 // activation is reduced to 0
(14) 𝑜 𝑖 = 1 − 𝑁 𝑛 𝜖 // output rises to 1
(15) // some activation is distributed to conn. neurons
(16) if ( 𝑗 is connected to 𝑖 via open gap junction)
(17) 𝑎 𝑗 ∶ = 𝑎 𝑗 + 𝜖
(18) }
(19) // temporal averaging of own output
(20) ̃ 𝑎 𝑖 = ( 1 − 𝛼 𝑡 ) ̃ 𝑎 𝑖 + 𝛼 𝑡 𝑜 𝑖
(21) // spatial averaging of temporal average
(22) 𝑎 𝑖 = 1 / ( 1 + 𝑁 𝑛 ) ( 𝑎 𝑖 + ∑ 𝑗 𝑎 𝑗 )
(23) 𝑎 𝑖 = ( 1 − 𝛼 𝑠 ) 𝑎 𝑖 + 𝛼 𝑠 ̃ 𝑎 𝑖
(24) // check if temporal average is above sync-threshold
(25) i f ( ̃ 𝑎 𝑖 > 𝑎 𝑖 )
(26) open gap junctions
(27) else
(28) close gap junctions