Delta learning rule it is derived from gradient descent method it can be generalized to more than one layer Updates weights between connections so as to minimize the difference between the net output and target value. Aim is to minimize the error overall training patterns Delta learning rule to update weights for ith pattern is Winew = Wiold + α(t-Yin)Xi Single neuron delta learning network is Adaline where as multi neuron delta learning network is Madaline. Inputs will be +1 or -1 where as output will be sign. Training algorithm (convert same to flowchart) 1. set initial weight values W, bias b and learning rate α 2. specify the error tolerance Es 3. for each Sensory input S and respective desired output t a) Xi = Si for i = 1 to n b) Calculate Yin = b + Σ XiWi c) Update the weights and bais as Winew = Wiold + α.(t-Yin).Xi and bnew = bold + α(tYin) d) Calculate error Ei = Σ(t-Yin)2 e) If Ei ≠ Es then repeat from step 3 else stop
© Copyright 2026 Paperzz