ASSIGNMENT 2

Artificial Intelligence
Section : 2A

Name : Kirthivasan a/l Sababathy

( SW090966)
Ashyok a/l Arumadurai
(SW089809)
Yugendramannan a/l Thiagarajoo
(SW090399)
Jagethes Rao a/l Nagarajan
(SW090965)

Question 1

a)
Genetic algorithm means to solve a problem with a set of instructions. These algorithms are
derivative-free, which means that they do not need routine al derivative information to search for
a set of parameters that minimize (or maximize) a given target function. Instead, they exclusively
rely on repeated valuation of the objective function, and the subsequent search focusing after
each evaluation follows certain heuristic rule of thumb. In particular, the optimum solution is
obtained by investigation new solutions which incorporate three genetic operations:
reproduction, crossing over , and mutation in a selective environs where the fittest survive. There
are three main steps for genetic algorithm, initialization of population (usually random),
evaluation of fitness office and finally generation of new population.
The individual are selected among the universe depending on their seaworthiness 6 senses of
value i.e. soul with lower fitness values are eliminated, whilst the others with higher fitness
values are copied to the next propagation one or more times in reproductive memory stage. The
population after reproduction stage is called mating pool.
Genetic crosswalk manipulator is applied to the mating pool to generate new individuals in
crossing microscope stage. Thus individuals of the mating pool are spanned randomly, and
genetic couples are obtained. There are many crossing wheeler dealer can be used but the most
staple crossover operator is the one-pointedness crossover operator, in this case a crossover point
in the string spot of the selected brace is randomly chosen, and the piece of the two parent are
interchanged at this point. In two-point crossover operation, the two crossover points are selected
in the binary program train of the pair under consideration and between these points the morsel
are swapped. This crossover process is similar to the mating process in a biological system,
where parent pass segments of chromosome to their offspring and thus offspring can outperform
their parents if they get goodness genes from both parents.

In a genetic mutation cognitive operation, this introduces further changes to a sting string. This is
required, since if the population does not contain all the encoded information required to solve a
specific problem, no sum of gene mix can provide a satisfactory result. By applying the mutation
operator, it is possible to produce new chromosome. This can be implemented in various
agencies, and the greenest proficiency is to change a randomly chosen bit in the bit string of the
individual to be mutated. Thus certain bit is changed from 1 into 0 or from 0 into 1.

(b)

Crossover
- Selects genes from parent chromosomes and creates a new offspring
- The crossover is done so that the new chromosome will likely create a better
individual where it takes the best characteristics from each of the parent.
- There are 4 types of crossover, namely:
i)
Single point crossover
ii)
Two point crossover
iii)
Uniform crossover
iv)
Arithmetic crossover
The crossover is like the figure below.
Chromosome 1
Chromosome 2
Offspring 1
Offspring 2

11011 | 00100110110
11011 | 11000011110
11011 | 11000011110
11011 | 00100110110

The crossover can be more difficult and depends on the encoding of the encoding
chromosomes.

Mutation
- After crossover is done, mutation takes place
- Mutation changes randomly to the new offspring
- For binary encoding, can randomly choose 0 to 1 , or 1 to 0
The mutation is like as figure below.

Original offspring 1
Original offspring 2
Mutated offspring 1
Mutated offspring 2

1101111000011110
1101100100110110
1100111000011110
1101101100111010

When the permutations are encoded, mutation could be exchanging the two genes.
(c)
Basically fitness routine determines which possible answers get passed on to multiply and mutate
into the next gene of root, and which will die out. Fitness function evaluate the effect and
reappearance a tantrum ness value (higher the figure, better the solution), rather than looking at
the attribute. The seaworthiness function looks into how good/optimal is the quality of the
proposed solution. Genetic Algorithm gives the attribute of the fit gene and the fitness function
should be able to determine if one gene is more fit than the other.

A common metaphor for the selection process to choose which potential solutions will

continue is of a large line roulette cycle.

The fitness values are then used in an outgrowth of natural survival to choose which
voltage solution will continue on to the next generation, and which will die out. It should
be noted, however, that natural option process does not merely choose the top x number
of solvent; the solutions are instead Chosen statistically such that it is more likely that a
solution with a higher seaworthiness value will be chosen, but it is not guaranteed. This
tends to correspond to the natural reality.

Question 2

Despite all different types of Neural Network that have been proposed, there are still considerate
amount of classic networks which have been widely used and on which many others are based
i.e. Hopfield Network, Back Propagation, Competitive Networks and networks using Spiky
Neurons.
Back propagation is training and learning algorithm rather the network itself. A Back
Propagation network learns by example. You give the algorithm examples of what you want the
network to do and it changes the networks weights so that, when training is finished, it will give
you the required output for a particular input. Back Propagation networks are ideal for simple
Pattern Recognition and Mapping Tasks.

A back-propagation neural network is only practical in certain situations. Following are some
guidelines on when you should use another approach:

Can you write down a flow chart or a formula that accurately describes the problem? If
so, then stick with a traditional programming method.

Is there a simple piece of hardware or software that already does what you want? If so,
then the development time for a Neural Network might not be worth it.

Do you want the functionality to "evolve" in a direction that is not pre-defined? If so,
then consider using a Genetic Algorithm (that's another topic!).

Do you have an easy way to generate a significant number of input/output examples of

the desired behavior? If not, then you won't be able to train your Neural Network to do
anything.

Is the problem is very "discrete"? Can the correct answer be found in a look-up table of
reasonable size? A look-up table is much simpler and more accurate.

Is precise numeric output values required? Neural Networks are not good at giving
precise numeric answers.

So, here are some elements where back propagation might be a good start:

A large amount of input/output data is available, but you're not sure how to relate it to the
output.
The problem appears to have overwhelming complexity, but there is clearly a solution.
It is easy to create a number of examples of the correct behavior.
The solution to the problem may change over time, within the bounds of the given input
and output parameters (i.e., today 2+2=4, but in the future we may find that 2+2=3.8).
Outputs can be "fuzzy", or non-numeric.