I Introduction to Neural Networks

Rule Based Approach

Rule Based Approach Refers to the AI modelling
where the relationship or patterns in data are
defined by the developer. The machine follows
the rules or instructions mentioned by the
developer, and performs its task accordingly.
For example, suppose you have a dataset
comprising of 100 images of apples and 100
images of bananas. To train your machine, you
feed this data into the machine and label each
image as either apple or banana. Now if you test
the machine with the image of an apple, it will
compare the image with the trained data and
according to the labels of trained images, it will
identify the test image as an apple. This is
known as Rule based approach. The rules given
to the machine in this example are the labels
given to the machine for each image in the
training dataset.

Learning Based Approach

Learning Based Approach Refers to the AI modelling where the relationship or patterns in data are
not defined by the developer. In this approach, random data is fed to the machine and it is left on
the machine to figure out patterns and trends out of it. Generally this approach is followed when
the data is unlabelled and too random for a human to make sense out of it. Thus, the machine
looks at the data, tries to extract similar features out of it and clusters same datasets together. In
the end as output, the machine tells us about the trends which it observed in the training data.
For example, suppose you have a dataset of 1000 images of random stray dogs of your area. Now
you do not have any clue as to what trend is being followed in this dataset as you don’t know their
breed, or colour or any other feature. Thus, you would put this into a learning approach based AI
machine and the machine would come up with various patterns it has observed in the features of
these 1000 images. It might cluster the data on the basis of colour, size, fur style, etc. It might
also come up with some very unusual clustering algorithm which you might not have even
thought of!

Decision Trees
It is a rule-based AI model which helps the machine in predicting what an element is with the help
of various decisions (or rules) fed to it. A basic structure of decision tree is shown below:

Here, the Decision tree starts from the question Am I Hungry? The beginning point of any
Decision Tree is known as its Root. It then forks into two different ways or conditions: Yes or No.
The forks or diversions are known as Branches of the tree. The branches either lead to another
question, or they lead to a decision like Go to Sleep which is known as the leaf. If you look closely
at the image above, you would notice that it looks like an inverted tree with root above and the
leaves below. Hence the name Decision Tree!
Points to Remember
● While making Decision Trees, one should take a good look at the dataset given to them and try
to figure out what pattern does the output leaf follow. Try selecting any one output and on its
basis, find out the common links which all the similar outputs have.
● Many times, the dataset might contain redundant data which does not hold any value while
creating a decision tree. Hence, it is necessary that you note down which are the parameters that
affect the output directly and should use only those while creating a decision tree.
● There might be multiple decision trees which lead to correct prediction for a single dataset. The
one which is the simplest should be chosen as the best.

Some of the common AI models are given below:

1. REGRESSION: This is a type of Rule based AI model. In regression,
the algorithm generates a mapping function from the given data,
represented by the solid line. The blue dots shown in the graph are the
data values and the solid line here represents the mapping done for
them. With the help of this mapping function, we can predict the future
data. For example, if we want to predict the salary of an employee, we
can use his past salaries as training data and can predict his next
salary. Regression works with continuous data.

2. CLASSIFICATION: This is another Rule based AI model. In

classification, the algorithm is able to determine which set a given
data point belongs to by means of a classification function
represented by the dotted line. The model classifies datasets
according to the rules given to it. Usually the dataset used for
classification are labelled and the data then gets sorted according to
their labelling. Testing data is then classified as one of the labels of
the training dataset. For example, If we want to train a model to
identify if an image is of a guitar or a piano, we need to train it with
multiple images of both guitar and piano along with their labels. The
machine will then classify images on the basis of the labels and
predict the correct label for testing data. Classification works on
discrete dataset

3. CLUSTERING: This is a Machine learning approach where the

machine generates its own rules or algorithms to differentiate amongst
the given dataset to achieve the pre-decided goal. The data fed to such
a model is usually unlabelled or random and thus the developer feeds
in the data directly into the machine and instructs it to build its own
algorithm. The machine then finds out patterns or trends out of the
training dataset and clusters the ones which follow the same pattern.
The output rules might be very different to what was expected as the
machine has its own way of recognising patterns. For example, if you
have a random data of stray dogs which live in your locality, since you
are unable to find any meaningful pattern amongst them, you would
feed their data into the clustering algorithm. The algorithm would
then analyse the data and divide them into clusters according to their similarities based on the
trends noticed. The clusters are then given as the output. Clustering works on discrete dataset.
 Neural Networks – Defined
A Neural Networks is a combination of algorithms to recognize underlying relationships in a set of
data which is like a process that mimics the way the human brain operates. Neural networks, also
known as artificial neural networks (ANNs).
Neural networks are loosely modelled after how neurons in the human brain behave. The key
advantage of neural networks, are that they are able to extract data features automatically
without needing the input of the programmer. A neural network is essentially a system of
organizing machine learning algorithms to perform certain tasks. It is a fast and efficient way to
solve problems for which the dataset is very large, such as in images.

Applications of neural networks

Some of the applications of neural networks are :

• Speech recognition
• Character recognition
• Spell checking

Speech recognition

Speech recognition has many applications such as virtual assistance , hands-free

computing and video games ,etc neural networks is used widely in this areas .

Character recognition

In this Application of AI is also a use of neural networks in character recognition, where

this technology is used to recognize a character with the help of this application of AI.

Spell Checking

Personalized Spell Checking using Neural Networks a new system for detecting misspelled
words was proposed. this system corrects the mistakes made by typist.

Some of the features of a Neural Network are listed below:

How Neural Networks work

In the above image as we can see that the Neural Networks is divided into 3
components.

Input layer In this layer, we insert data.

Multiple hidden Invisible layer which processes all the input

layer data.

Output layer Gives output of data that was inserted.

1. In the first layer of Neural Networks , there is input data in which we inserts
data.
2. In the second layer of Neural Networks , there is multiple hidden layers are there
which are not visible but all processing occurs in these layers.
3. In the third layer of Neural Networks , there is output data which gives the
output of data which was processed in the hidden layers.

Features of Neural Networks(ANN)

• Neural Networks(NN) system are modelled on the functioning of human
brain.
• They are able to automatically extract features without input from the
programmer.
• Every neural network is essentially a machine learning algorithm.
• It is useful when solving problems for which the data set is very large.
Neural networks vs Human Nervous system

Given are the images of a Human Neuron and its relation with the Neural Network. The axon
from a neuron sends an impulse to another neuron. The impulse received is then sent to the cell
body (nucleus) through dendrites. The cell body performs an activation function on the impulse
received and then gives it to the output axon which passes the same to the next neuron in the
system. Now as we relate this process with an Artificial Neural Network, we can see that the
input layer gets data which is passes on to the nodes in the hidden layer. The nodes perform
specific actions on the data and pass the processed information to the next layer. In the end, the
final processed data reaches the output of the system.

How neural networks learn!

There are three ways of neural networks learning:

1. Supervised Learning
2. Unsupervised Learning
3. Reinforcement Learning

Supervised Learning
As the name supervised learning indicates there is a supervisor who teaches or trains
the model with respect to the output. In this type of learning, networks are trained to
provide the correct output by using several example inputs. For example, the network
can be trained on the data of students and marks. After training the network will be able
to calculate the marks based on various elements like the length of the answer, correct
facts written in the answers, comprehension length etc.

Unsupervised Learning
In this learning no output related information received by the neural network and it
provides only input. The system evaluates and find out how different elements are
related. This method used for solving clustering problems, estimation problems, self-
organizing maps etc.

Reinforcement Learning
Reinforcement learning, the neural network learn based on learning from experience.
Unlike supervised learning, reinforcement learning doesn’t have any output information.
It is employed by various software and machines to find the best possible behavior or
path it should take in a specific situation.