Skip to main content

Table 1 GHA-basedspike feature extraction

From: Stream-based Hebbian eigenfilter for real-time neuronal spike discrimination

Input:

Neuronal spikes, x → (i) ;

Initial synaptic weight, W → (1) ;

Learning rate, η ;

Output:

Principal components of neuronal spikes, W → (N) ;

1. Initialize synaptic weight W → (1) and learning rate η, j = 1

2. Calculate the mean vector of the aligned spikes

    μ → = ∑ i = 1 n x → (i)/n

3. Zero-mean transformation

    x → (i) = x → ( i ) - μ → 1≤ i ≤ n

4. Perform Hebbian learning on zero-mean data

    y → (j) = W → (j) x → (i)

    LT(j) = LT y → (j) y → T (j)

    dW(j) =η y → (j) x → T ( i ) - LT(j) W → (j)

    W → (j + 1) = W → (j) + dW(j)

5. If j= N, the algorithm stops, otherwise j = j+ 1, go to step 4