Perceptron algorithm in pseudo-code

Start with random initial weights (e.g., uniform random in [-.3,.3])

Do

{

For All Patterns p

{

For All Output Nodes j

{

CalculateActivation(j)

Error_j = TargetValue_j_for_Pattern_p - Activation_j

For All Input Nodes i To Output Node j

{

DeltaWeight = LearningConstant * Error_j * Activation_i

Weight = Weight + DeltaWeight

}

}

}

}

Until "Error is sufficiently small" Or "Time-out"