EL40102 ARTIFICIAL INTELLIGENCE

7th EEE
Assignment – 1

1) Biological neurons have a cell body, axons, dendrites and synapses. Draw a diagram
and label these terms on it. Draw the computational equivalent and label it.
2) Derive expressions for the weights and threshold of a Perceptron that computes the
logical function OR.
3) When a BP algorithm is used, the error function must be differentiable. Why? In a feed-
forward neural network trained by BP, what is back-propagated, and from where to
where? In a feed-forward network using a sigmoidal (logistic) activation function, why
might one avoid setting target output values 0.0 and 1.0, and prefer to use 0.1 and 0.9
(or even 0.2 and 0.8)
4) What does the learning rate do in back-propagation training? What may happen when
we use a very large learning rate, and when we use a very small learning rate?
5) Discuss the similarities and differences between back-propagation (BP) and radial basis
function (RBF) networks.
6) Fig. 1 shows the diagram of an artificial neuron.
The weights w1 = 2, w2 = – 4, w3 = 1, and the
activation unit is given by step function:
1 𝑖𝑓 𝑣 ≥ 0
𝜑 (𝑣 ) =
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Calculate what will be the output value y of the unit
for each of the following input patterns:
Fig. 1 Artificial Neuron
Pattern P1 P2 P3 P4
X1 1 0 1 1
X2 0 1 0 1
X3 0 1 1 1