Introduction

to Neural
Networks
M Yogi Reddy
Assistant Professor
CSE Department
School Of Technology
GITAM(Deemed to be ) University
Hyderabad

PAGE 1
Agenda

 Introduction to Neural Networks

 Deep Learning
 Reinforcement Learning

PAGE 2
 Introduction to Neural
Networks
 A method of computing, based on the interaction of multiple connected processing
elements.
 A powerful technique to solve many real world problems.
 The ability to learn from experience in order to improve their performance.
 Ability to deal with incomplete information
 Basics Of Neural Network
1. Biological approach to AI
2. Developed in 1943
3. Comprised of one or more layers of neurons
4. Several types, we will focus on feed-forward and feedback
networks
PAGE 3
 Neurons

Biological Artificial
PAGE 4
Neural Network Neurons

• Receives n-inputs
• Multiplies each input by its
weight
• Applies activation function to
the sum of results
• Outputs result

PAGE 5
 Activation Functions

• Controls when unit is

“active” or “inactive”
• Threshold function outputs
1 when input is positive and 0
otherwise
• Sigmoid function
= 1 / (1 + e-x)
PAGE 6
Neural Network types can be classified based on following
attributes:
 Connection Type
- Static (feed-forward)
- Dynamic (feed-back)
 Topology
- Single layer
- Multilayer
- Recurrent
 Learning Methods
- Supervised
- Unsupervised
- Reinforcement PAGE 7
 Classification Based On Connection Types
Static(Feed-forward) Dynamic(Feedback)

The output is calculated directly The output depends also on the

from the input through Feed- previous inputs, outputs or States of
forward connections the network

PAGE 8
 Classification Based On Topology
Topology defines how a neuron in neural network connected with
another neurons.
There are three types topologies that every neural network must
follow the one of the following:
1. single-level topology
2. multi-level topology
3. recurrent topology
1. single-level topology:
The simplest kind of neural network is a single-layer
network, which consists of equal no.of input and
output nodes.

PAGE 9
2. multi-level topology
In multi-level, each neuron in one layer has directed connections to the neurons of the subsequent layer

3. recurrent topology
A recurrent neural network (RNN) is a class of artificial neural networks where connections between units form a directed cycles.

PAGE 10
 Learning methods of Neuron:
Neurons in neural networks will learn about the working pattern of the
new task.
Next time, when the same task is given to perform, it automatically
generates output without wasting of time.
There are three types of learning methods. they are
1. supervised learning
2. unsupervised learning
3. reinforcement learning

PAGE 11
 Neural Network Applications
 Pattern recognition
 Investment analysis
 Control systems & monitoring
 Mobile computing
 Marketing and financial applications
 Forecasting – sales, market research, meteorology

PAGE 12
 Advantages: Disadvantages:
•A neural network can perform tasks that a •The neural network needs training
linear program can not. to operate.
•When an element of the neural network •The architecture of a neural
fails, it can continue without any problem network is different from the
by their parallel nature. architecture of microprocessors
•A neural network learns and does not therefore needs to be emulated.
need to be reprogrammed. •Requires high processing time for
•It can be implemented in any application. large neural networks.
•It can be implemented without any
problem

PAGE 13
Deep Learning
 Deep learning is a branch of machine learning which is completely based
on artificial neural networks, as neural network is going to mimic the human
brain so deep learning is also a kind of mimic of human brain.
 In deep learning, we don’t need to explicitly program everything.
 It’s on hype nowadays because earlier we did not have that much processing
power and a lot of data.
 As in the last 20 years, the processing power increases exponentially, deep
learning and machine learning came in the picture.
 A formal definition of deep learning is- neurons
“Deep learning is a particular kind of machine learning that achieves great power
and flexibility by learning to represent the world as a nested hierarchy of concepts,
with each concept defined in relation to simpler concepts, and more abstract
representations computed in terms of less abstract ones”.
PAGE 14
 Architectures:

Deep Neural Network – It is a neural

network with a certain level of
complexity (having multiple hidden
layers in between input and output
layers). They are capable of modelling
and processing non-linear
relationships.
Deep Belief Network(DBN) – It is a
class of Deep Neural Network. It is
multi-layer belief networks.

PAGE 15
Advantages:
Best in-class performance on problems.
Reduces need for feature engineering.
Eliminates unnecessary costs.
Identifies defects easily that are difficult to detect.
Disadvantages:
Large amount of data required.
Computationally expensive to train.
No strong theoretical foundation.

PAGE 16
Applications:
 Automatic Text Generation – Corpus of text is learned and from this model new
text is generated, word-by-word or character-by-character.
Then this model is capable of learning how to spell, punctuate, form sentences,
or it may even capture the style.

 Healthcare – Helps in diagnosing various diseases and treating it.

 Automatic Machine Translation – Certain words, sentences or phrases in one

language is transformed into another language (Deep Learning is achieving top
results in the areas of text, images).

 Image Recognition – Recognizes and identifies peoples and objects in images as

well as to understand content and context. This area is already being used in
Gaming, Retail, Tourism, etc. PAGE 17
 Reinforcement Learning

 A reinforcement learning algorithm, or

agent, learns by interacting with its
environment.
 The agent receives rewards by performing
correctly and penalties for performing
incorrectly. The agent learns without
intervention from a human by maximizing
its reward and minimizing its penalty.
 It is a type of dynamic programming that
trains algorithms using a system of reward
and punishment.

PAGE 18
 Given example, we can see that the agent
is given 2 options i.e. a path with water or
a path with fire.
 A reinforcement algorithm works on
reward a system i.e. if the agent uses the
fire path then the rewards are subtracted
and agent tries to learn that it should
avoid the fire path.
 If it had chosen the water path or the safe
path then some points would have been
added to the reward points, the agent
then would try to learn what path is safe
and what path isn’t.

PAGE 19
PAGE 20