HINDUSTHAN COLLEGE OF ENGINEERING & TECHNOLOGY
21CA3206 – DEEP LEARNING TECHNIQUES
(Autonomous – R2020)
UNIT - 1 - BASICS
Deep Learning – History - Key facts – Biological Neuron- Artificial Neuron – Mc Culloch
Pitts Neuron, Limitations of MP Neuron – Perceptron – Multi-Layer Perceptron – Basics of
Neural Networks :Neurons, Output Functions – Activation Functions – Weights and Biases
-----------------------------------------------------------------------------------------------------------------
Deep Learning – History - Key facts – Biological Neuron- Artificial Neuron – Mc
Culloch Pitts Neuron, Limitations of MP Neuron – Perceptron – Multi-Layer
Perceptron – Basics of Neural Networks :Neurons, Output Functions – Activation
Functions – Weights and Biases
Table of Contents
NEURAL NETWORKS ........................................................................................... 1
OBJECTIVE OF ARTIFICIAL NEURAL NETWORK .............................................. 2
PROPERTIES OF NEURAL NETWORK ................................................................. 4
APPLICATIONS OF NEURAL NETWORKS........................................................... 6
BIOLOGICAL NEURON ......................................................................................... 8
INSPIRATION TO ARTIFICIAL NEURON (PERCEPTRON): ................................. 9
NEURAL NETWORKS
A neural network is a processing system
It can be either in the form of hardware device or software algorithm or both .
The design of neural network is inspired by the design and functioning of human
brain.
A neural network can be defined as a huge distributed information processing system
made up of simpler interconnected units called as neurons or nodes which work
together collectively.
It replicates the functioning of human brain.
UNIT – 1 Page No : 1 Prepared by : M.Nirmala / AP / MCA / HICET
, HINDUSTHAN COLLEGE OF ENGINEERING & TECHNOLOGY
21CA3206 – DEEP LEARNING TECHNIQUES
(Autonomous – R2020)
UNIT - 1 - BASICS
Deep Learning – History - Key facts – Biological Neuron- Artificial Neuron – Mc Culloch
Pitts Neuron, Limitations of MP Neuron – Perceptron – Multi-Layer Perceptron – Basics of
Neural Networks :Neurons, Output Functions – Activation Functions – Weights and Biases
-----------------------------------------------------------------------------------------------------------------
Neural Networks are processing systems, the design which is inspired by the
structure and functioning of human brain.
OBJECTIVE OF ARTIFICIAL NEURAL NETWORK
APPLICATIO LEARNING /
N TRAINING
GENERALIZA
TION /
RECOGNITIO
N
UNIT – 1 Page No : 2 Prepared by : M.Nirmala / AP / MCA / HICET
, HINDUSTHAN COLLEGE OF ENGINEERING & TECHNOLOGY
21CA3206 – DEEP LEARNING TECHNIQUES
(Autonomous – R2020)
UNIT - 1 - BASICS
Deep Learning – History - Key facts – Biological Neuron- Artificial Neuron – Mc Culloch
Pitts Neuron, Limitations of MP Neuron – Perceptron – Multi-Layer Perceptron – Basics of
Neural Networks :Neurons, Output Functions – Activation Functions – Weights and Biases
-----------------------------------------------------------------------------------------------------------------
The main goal of Artificial Neural Networks (ANNs) is to help computers learn and
make decisions by imitating how our brains work. They're like virtual brains that
can understand patterns in data, learn from examples, and make smart choices.
Predicting whether you need to carry an umbrella based on weather conditions.
Imagine you want to build a simple computer program that can tell you whether
you should take an umbrella with you when you go outside. Instead of giving it a
list of rules like "carry an umbrella if it's raining" or "don't carry an umbrella if
it's sunny," you want the computer to learn by itself.
Setting Up:
An Artificial Neural Network called as Mini brain is setup.
The brain network will learn when to recommend carrying an umbrella.
Learning:
Different learning mechanisms are given to the brain to learn by itself like
Some days with Rain
Some days without rain
The brain is educated by giving the information whether the umbrella will be carried
during rainy days or not. Once the information is provided, the network starts
noticing the patterns like the
Sky looks (bright / dark)
Temperature
Humidity
Making Predictions:
After learning from many examples, you give the network a new day's information –
how the sky looks, the temperature, and humidity. The network uses what it learned
to make a guess about whether you should carry an umbrella or not.
Feedback: If the network guesses wrong, you correct it. Over time, it adjusts itself
based on the feedback you provide, so it gets better at predicting whether you need
an umbrella.
UNIT – 1 Page No : 3 Prepared by : M.Nirmala / AP / MCA / HICET
, HINDUSTHAN COLLEGE OF ENGINEERING & TECHNOLOGY
21CA3206 – DEEP LEARNING TECHNIQUES
(Autonomous – R2020)
UNIT - 1 - BASICS
Deep Learning – History - Key facts – Biological Neuron- Artificial Neuron – Mc Culloch
Pitts Neuron, Limitations of MP Neuron – Perceptron – Multi-Layer Perceptron – Basics of
Neural Networks :Neurons, Output Functions – Activation Functions – Weights and Biases
-----------------------------------------------------------------------------------------------------------------
PROPERTIES OF NEURAL NETWORK
NON LINEARITY
It captures relationship between non linear data.
Helps to capture relationship between complex input and output data which are not
expressed using linear form.
SELF ORGANIZATION
It can create the or own organization or representation during learning time
LEARNING FROM DATA
Neural networks can learn from examples and data. They adjust their internal
parameters (weights and biases) based on the patterns present in the training data,
enabling them to make predictions or classifications on new, unseen data.
Email Spam Detection to illustrate the concept of learning from data using
neural network
Assume you have labelled emails and create a neural network
Steps are
Data Collection
Large dataset which comprises information of emails details. , some of which are
labeled as "spam" and others as "not spam" (ham). Each email is represented as a
collection of features like the frequency of certain words, presence of link s, and other
relevant characteristics.
Sample Email Spam Detection Dataset:
Email Text Label (Spam/Not Spam)
Hey there! Get the best deals on our website today! Spam
Congratulations, you've won a prize! Spam
UNIT – 1 Page No : 4 Prepared by : M.Nirmala / AP / MCA / HICET
21CA3206 – DEEP LEARNING TECHNIQUES
(Autonomous – R2020)
UNIT - 1 - BASICS
Deep Learning – History - Key facts – Biological Neuron- Artificial Neuron – Mc Culloch
Pitts Neuron, Limitations of MP Neuron – Perceptron – Multi-Layer Perceptron – Basics of
Neural Networks :Neurons, Output Functions – Activation Functions – Weights and Biases
-----------------------------------------------------------------------------------------------------------------
Deep Learning – History - Key facts – Biological Neuron- Artificial Neuron – Mc
Culloch Pitts Neuron, Limitations of MP Neuron – Perceptron – Multi-Layer
Perceptron – Basics of Neural Networks :Neurons, Output Functions – Activation
Functions – Weights and Biases
Table of Contents
NEURAL NETWORKS ........................................................................................... 1
OBJECTIVE OF ARTIFICIAL NEURAL NETWORK .............................................. 2
PROPERTIES OF NEURAL NETWORK ................................................................. 4
APPLICATIONS OF NEURAL NETWORKS........................................................... 6
BIOLOGICAL NEURON ......................................................................................... 8
INSPIRATION TO ARTIFICIAL NEURON (PERCEPTRON): ................................. 9
NEURAL NETWORKS
A neural network is a processing system
It can be either in the form of hardware device or software algorithm or both .
The design of neural network is inspired by the design and functioning of human
brain.
A neural network can be defined as a huge distributed information processing system
made up of simpler interconnected units called as neurons or nodes which work
together collectively.
It replicates the functioning of human brain.
UNIT – 1 Page No : 1 Prepared by : M.Nirmala / AP / MCA / HICET
, HINDUSTHAN COLLEGE OF ENGINEERING & TECHNOLOGY
21CA3206 – DEEP LEARNING TECHNIQUES
(Autonomous – R2020)
UNIT - 1 - BASICS
Deep Learning – History - Key facts – Biological Neuron- Artificial Neuron – Mc Culloch
Pitts Neuron, Limitations of MP Neuron – Perceptron – Multi-Layer Perceptron – Basics of
Neural Networks :Neurons, Output Functions – Activation Functions – Weights and Biases
-----------------------------------------------------------------------------------------------------------------
Neural Networks are processing systems, the design which is inspired by the
structure and functioning of human brain.
OBJECTIVE OF ARTIFICIAL NEURAL NETWORK
APPLICATIO LEARNING /
N TRAINING
GENERALIZA
TION /
RECOGNITIO
N
UNIT – 1 Page No : 2 Prepared by : M.Nirmala / AP / MCA / HICET
, HINDUSTHAN COLLEGE OF ENGINEERING & TECHNOLOGY
21CA3206 – DEEP LEARNING TECHNIQUES
(Autonomous – R2020)
UNIT - 1 - BASICS
Deep Learning – History - Key facts – Biological Neuron- Artificial Neuron – Mc Culloch
Pitts Neuron, Limitations of MP Neuron – Perceptron – Multi-Layer Perceptron – Basics of
Neural Networks :Neurons, Output Functions – Activation Functions – Weights and Biases
-----------------------------------------------------------------------------------------------------------------
The main goal of Artificial Neural Networks (ANNs) is to help computers learn and
make decisions by imitating how our brains work. They're like virtual brains that
can understand patterns in data, learn from examples, and make smart choices.
Predicting whether you need to carry an umbrella based on weather conditions.
Imagine you want to build a simple computer program that can tell you whether
you should take an umbrella with you when you go outside. Instead of giving it a
list of rules like "carry an umbrella if it's raining" or "don't carry an umbrella if
it's sunny," you want the computer to learn by itself.
Setting Up:
An Artificial Neural Network called as Mini brain is setup.
The brain network will learn when to recommend carrying an umbrella.
Learning:
Different learning mechanisms are given to the brain to learn by itself like
Some days with Rain
Some days without rain
The brain is educated by giving the information whether the umbrella will be carried
during rainy days or not. Once the information is provided, the network starts
noticing the patterns like the
Sky looks (bright / dark)
Temperature
Humidity
Making Predictions:
After learning from many examples, you give the network a new day's information –
how the sky looks, the temperature, and humidity. The network uses what it learned
to make a guess about whether you should carry an umbrella or not.
Feedback: If the network guesses wrong, you correct it. Over time, it adjusts itself
based on the feedback you provide, so it gets better at predicting whether you need
an umbrella.
UNIT – 1 Page No : 3 Prepared by : M.Nirmala / AP / MCA / HICET
, HINDUSTHAN COLLEGE OF ENGINEERING & TECHNOLOGY
21CA3206 – DEEP LEARNING TECHNIQUES
(Autonomous – R2020)
UNIT - 1 - BASICS
Deep Learning – History - Key facts – Biological Neuron- Artificial Neuron – Mc Culloch
Pitts Neuron, Limitations of MP Neuron – Perceptron – Multi-Layer Perceptron – Basics of
Neural Networks :Neurons, Output Functions – Activation Functions – Weights and Biases
-----------------------------------------------------------------------------------------------------------------
PROPERTIES OF NEURAL NETWORK
NON LINEARITY
It captures relationship between non linear data.
Helps to capture relationship between complex input and output data which are not
expressed using linear form.
SELF ORGANIZATION
It can create the or own organization or representation during learning time
LEARNING FROM DATA
Neural networks can learn from examples and data. They adjust their internal
parameters (weights and biases) based on the patterns present in the training data,
enabling them to make predictions or classifications on new, unseen data.
Email Spam Detection to illustrate the concept of learning from data using
neural network
Assume you have labelled emails and create a neural network
Steps are
Data Collection
Large dataset which comprises information of emails details. , some of which are
labeled as "spam" and others as "not spam" (ham). Each email is represented as a
collection of features like the frequency of certain words, presence of link s, and other
relevant characteristics.
Sample Email Spam Detection Dataset:
Email Text Label (Spam/Not Spam)
Hey there! Get the best deals on our website today! Spam
Congratulations, you've won a prize! Spam
UNIT – 1 Page No : 4 Prepared by : M.Nirmala / AP / MCA / HICET
