You are on page 1of 4

1. Define depolarization?

A: The membrane of the neuron is permeable to certain ionic species, and acts to maintain a potential difference between the intracellular fluid and the extracellular fluid. Influx of positive species into the cell tends to depolarize the resting potential. In depolarization, sodium ions penetrate into the membrane and potassium ions penetrate out of the membrane in order to respond for the excitatory inputs and to reduce the potential across the membrane.

2. What are the 3 different layers in an ANN?


A: The common type of artificial neural network consists of three groups, or layers, of units: a layer of "input" units is connected to a layer of "hidden" units, which is connected to a layerof "output" units.

3. What is competitive learning?


A:Competitive learning is the method in which the neurons that respond strongly to the input stimuli get their weights updated. When an input pattern is presented, all neurons in the layer compete and the winning neuron gets its weight updated. This is known as the WINNERTAKES- ALL.

4. Name any 2 unsupervised learning algorithms?


A:The two unsupervised learning algorithms are Hebbian learning and Competitive learning algorithms.

5. What are the advantages of supervised learning? A: With supervised learning the network has a set of training data and the output is explicit. It uses the data to adjust the networks weights and threshold so as to minimize the error in its predictions on the training set. If the network is trained properly it has then learned to model the function that relates the input variables and can subsequently be used to make predictions where output is unknown. 6. Advantage of bipolar data over binary data? A: In case of bipolar vectors the missing and mistake vector in the pattern can be differentiated easily. The main advantage of bipolar coding over binary coding is that bipolar vectors have a greater probability of being orthogonal than binary vectors. This will be an issue when we start storing vectors in the associative networks. 7. Difference between bias and threshold? A: Bias: Bias is a weight on a connection from a unit whose activation is always one(1). Threshold: Threshold is the limit, a value which is equal or greater than this is considered to be action potential otherwise it is considered to be resting potential. It is denoted by the symbol theta. The sum of the weighted input signal (i.e products of the weights and the inputs) is calculated in each node, and if the value is above some threshold (typically 0) the neuron fires and takes the activated value (typically 1); otherwise it takes the deactivated value (typically -1 or 0).

8. Give the equation for bipolar sigmoidal function and its significance in ANN. A: Bipolar sigmoidal function is

f (x) =

The main significance is that the system has a desired range of output values from -1 to +1 and is closely related to hyperbolic tangent functions and provides ease of computation. 9. Define feedforward and recurrent network?
A: Feed forward: In the feedforward neural network the information moves in only one direction, forward, from the input nodes, through the hidden nodes (if any) and to the output nodes. There are no cycles or loops in the network. Recurrent: A recurrent neural network (RNN) is a class of neural network where connections between units form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior. Unlike feedforward neural networks, RNNs can use their internal memory to process arbitrary sequences of inputs. This makes them applicable to tasks such as unsegmented connected handwriting recognition, where they have achieved the best known results.

10. What is a multilayer network?


A: A multilayer network is a network with one or more layers (or levels) of nodes (so called hidden layers) between the input and the output layer. Typically there is a layer of weights between the two adjacent levels of units. Multilayer networks can solve more complicated problems than single layer network.

11. Draw the MC-Culloh pitt's network for N3(t)=N2(t-2)&


A:

N1(t-3)

12. Implement half adder using Mc-Culloch Pitts method. A 0 0 1 1 B 0 1 0 1 S C 0 0 1 0 1 0 0 1 S= A XOR B

C=A.B 13. What is a hebb net?

14. Why XOR can't be resolved by Hebb's rule? A: Hebbian learning is limited only for single layer network and the nature of problem need to be linearly separable. But for resolving XOR , it needs multi layer or hidden layer and it cant be linearly separable. So, XOR can't be resolved by Hebb's rule. 15. Draw the architecture of a single layer perceptron? A:

16. What are the limitations of perceptron? A: a) Hard limiting algorithm b)Linear separability 17. What are the main components of a BNN? A: Three main components of a biological neuron are the dendrites, the soma, the axon. Some more components are axon hillock, the myelin sheath, nodes of Ranvier and the synapse. 18. What are three important characteristics of an activation function in a BPN? A: Continuous Differentiable Monotonically non-decreasing 19. What are the difficulties encountered while training a BPN? A: The major problems are slow convergence and possible termination at a non-global solution. 20. How can the local minima be avoided? Occurrence of local minima in the criterion function is often related specific patterns of defects in classifier. Repeated training with random starting weights is the popular method to avoid this problem, but it requires extensive computational time. This problem is overcome with the help of Enhanced Back Propagation (EBP) algorithm. 21. What is a local minima in BPN? A: A local minima is a false optimal weight matrix that prevents the back propagation training algorithm from seeing the true solution.

You might also like