SlideShare a Scribd company logo
DEEP LEARNING
By
A. Yamini Saraswathi
Y18MTSP401
Under the guidance of
Mr. T. Krishna chaithanya (M.Tech)
Contents
• Artificial intelligence, machine learning
• What is deep learning
• computer vision features
• Fundamentals of deep learning
• Back propogation algorithm
• Convolutional neural network
• Recurrent neural network
• Deep learning framework
• Applications
Artificial intelligence
Artficial intelligence is the simulation of human intelligence
processes by machines especially computer systems. In this process
include learning, reasoning to reach to an approximate or accurate
output.
Continution....
Fig: Artificial neural network
Fig: Biological neuron
Continution....
Applications of artficial intelligence
• Autonomous vehicles(as drones and self driven cars)
• Medical diagnosis
• Playing games (such as chess)
• Online assistance (such as siri)
• Image recognition
• Speech recognition
• Natural language processing
• Etc......
Continution...
Learning of a neural netwwork can be of two types
• Supervised learning:-
It is used to perform classification when we want to map input
data to output labels. In this data is labeled with classes.
• Unsupervised learning:-
This learning is used to perform clustering and in this data is not
labeled. It is mostly used in exploratory analysis.
Machine learning
• Which is a subclass of artificial intelligence that automatially learn
and improve from the experience without being epxlicitly
programmed.
• Intelligent systems also built on machine learning algorithms which
have the capability to learn from the past experience.
Continution....
MACHINE
Trained data:
Rose
Tulips
Jasmine
Chrysanthymum
Input data
Feature
extraction
Chrysanthy
mum
Output
Limitations of machine learning
• High dimensionality of the data
• Unable to solve crucial AI problems
• One of the biggest challenge is feature extraciton
• Requires only significant amount of data
What is deep learning
Which is a subclass of
machine learning. In this neural
network is designed with more
than one hidden layer between
input and output layer. It reduces
the complexity of high dimensions
of the data.
Fig: Deep neural network
Computer vision features
• Camera obscura
• www.youtube.com/watch?v=10Hay06LJ4
• Black world
• SIFT feature
• Object recognition
• Etc......
Fundamentals of deep learning
Activation function:-
These are really important for artificial neural network to learn and
make sense of something really complicated and make non-linear complex
functional mapping between inputs and response variable. It should be
always non-linear and there are so many types
Sigmoid
Tanh
Softmax
Relu
Leaky Relu
Continution....
Sigmoid
It looks like an S- shaped curve. It's mathematical form is f(x) = 1 / 1 +
exp(-x) and its range is from 0 to 1 easy to understand and apply but it
has some major issues
Vanishing gradient
It's output is not zero centered
Sigmoids saturate and kill gradients
Sigmoids have slow convergence
Continution....
Tanh
It also mostly looked like S-shaped
curve and its range is from -1 to 1
and it's mathematical form
f(x) = 1-exp(-2x) / 1 + exp(-2x)
in this optimizations is easier. it's
major issue is
Vanishing gradient problem
Continution....
Relu
It has become more popular in last
couple of years. It was recently proved
that it had 6 times improvement in
convergence compared to Tanh
function it also solves vanishing
gradient problem there are two major
issues
• It can be used in hidden layers only
• Some gradients may die
Continution....
Leaky Relu
To fix the problem of dying neurons in the
Relu we will use Leaky Relu which is the
modification of Relu which introdues
a small slope to keep the updates alive
Loss function
• It is having an important part in ANN which is used to measure the
inconsistency between predicted value and actual value. These are
very helpful for training the neural network.
Y=Predicted output-actual output
• Many types of loss functions are present to deal with different types
of tasks but most widely used loss function is Mean Square
Error(MSE)
Optimization
Weights and bias are modifiedusing a function calledoptimization
function in such a way error is minimized. For updating weights we use an
equation
Optimazation algorithms are pf 2 types
• First order optmization algorithm
• Second order optimization algorithm
Optimzation algorithms
Gradient descent is the most popularly used optmization algorithm.
Which is of two types
• Stochastich gradient descent
θ=θ−η∇J(θ) where ∇J(θ) is the Gradient of Loss function J(θ) w.r.t
parameters 'θ’
• Mini batch gradient descent
Weight initialization
The method that we follow for weight initilalization is 'Xavier
intialization'. This method tries to intialize the weights with a smarter
value
Normalization
It is most important part in ANN because because most of the
data obtained from the real world are very distant each other. So to
keep the data at a comparable range we need to normalize the input
data of input layer. Common normalization approach includes.
• Statistical normalization
• Sigmoid normalization
Batch normalizaton
• Training deep neural network with more number of layers is more
challenging one to avoid this problem 'Batch normalization' is
introduced.
• It trains deep neural network that standardizes the inputs to a
layernfor each mini-batch and
Benifits of this batch normalization are
• Network train
• Makes weights easier to initialize
• Simplifies the creation of deeper network
• Makes more activation functions viable
Continution...Normalization
Batch normalization
Fig: Deep neural network
Regularization
When we build a neural network it performs well for trained data but
not on the test data. This is the sign of over fitting to reduce this
problem there are two ways
• Use more data
• Regularization
Getting more data is sometimes impossible and expoensive so for this
purpose we use 'Regularization'
So to solve this we should make our model generalize over the trained
data which is done using various regularization techniques.
Continuation...
Regularization techniques
• L1 & L2 regularization
Which updates general cost function by adding another term known as
regularization term
There is a small difference between L1 and L2 in the equation. Mostly L1 is
preferred
Continuation....
• Dropout regularization
This is mostly used one in the deep learning. At every iteration it randomly
selects some nodes and removes remaining with all of their incoming
signals and outgoing connections
So at each iteration has a different set of nodes
which result a different set of output
At last we ensemble models which usually perform
better than a normal neural network
Fig: Drop out Neural network
Back propogation algorithm
Neural network is trained by 'Back propogation algorithm' after
forwarding inputs to neural network then inputs and weights are summed
and multiplied then apply to activation function at last we predict the error
so to reduce this error we need to update weights using gradient descent
and bias which propogate backwwards to network
Convolutional neural network
In todays, convnets are used in every where it has been very effective
in such as image classification and image recognition it has been very
successful in identifying faces, objects and traffic signs
LeNet was one of the first convolutional neural network which helped
the path to deep learning
Convent is mainly build based on 3 layers
• Convolution layer & Non linearity
• Pooling layer
• Fully connected layer
Continution...
Fig: Convolutional neural network
Continution....
Convolution operation
Fig:Input image Fig: Kernal filter
Fig: Feature map
K=1
Continution....
Padding & pooling operation
Fig: Pooling operation
Fig: Padding operation
Continution....
Visulalisation of Convnet
Continuation...
Fig: Practical example of Convnet
Recurrent Neural Network
• RNN are also a feed forward neural network, however with ecurrent
meomory loops which takes input from the previous and /or same
layers of states
• This gives them a unique capability to model along the time
dimension and arbitary sequence of even inputs.
• RNN's are used for sequenced data analysis such as time series,
sentimental nalysis, NLP, language translation, speech recognition,
image captioning, script recognition and other othings.
• These are aslo calles networks with memory as the previous inputs or
states may persist in the model to do a sequential analysis. These
memories become input as well
Google Net
• It is almost different compared to other architectures like Alex Net
• It is the winner of Image Net 2014 and can train network faster which
is developed by Google
• In this global average pooling is used at the end of architecture it
leads to reduce number of weights are reduced and accuracy
increases to 0.6%
Continuation…
Deep Learning Framework
• Deep learning frameworks were built in academia in first generation Caffe
from Berkley university
Torch from NYU and also collaboration with Facebook
• In second generation all deep learning frameworks were developed from
industry's
Caffe2 from Facebook
Pytarch from Facebook
Tensor flow is from Facebook
MXNet from Amazon
CNTK form Microsoft
Importance of Deep learning frame work
• Easily can build big computational graphs
• Easily compute gradients in computational graphs
• Run it all efficiently n GPU
Continuation...
• This framework is an open source which aims to simplify the
implementation of compels and large scale deep learning models
using these amazing frameworks we can build complex models like
CNN. Some types of frameworks listed below.
Tensor flow
Keras
Pytorch
Caffe
Deeplearning4j
• One of the most popular frame work is 'Tensorflow'
Applications
• Automatic speech recognition
• Image recognition
• Visual art processing
• Natural language processing
• Drug discovery and toxicology
• Bioinformatics
• Medical Image Analysis
• Mobile advertising
• Image restoration
• Financial fraud detection
• Military
Google Arts & Culture app
This app uses automatic image analysis to compare lot of pictures in
the museums and in institutions throughout the world with your
picture. For this process mainly deep learning is mainly involved and in
the next slide I have taken the pictures practically.
Visualization of Deep Learning
Visualization of Deep Learning

More Related Content

What's hot

Convolutional neural network
Convolutional neural network Convolutional neural network
Convolutional neural network
Yan Xu
 
Overview of Convolutional Neural Networks
Overview of Convolutional Neural NetworksOverview of Convolutional Neural Networks
Overview of Convolutional Neural Networks
ananth
 
CNN and its applications by ketaki
CNN and its applications by ketakiCNN and its applications by ketaki
CNN and its applications by ketaki
Ketaki Patwari
 
Convolutional Neural Networks
Convolutional Neural NetworksConvolutional Neural Networks
Convolutional Neural Networks
Ashray Bhandare
 
Convolutional Neural Network (CNN) - image recognition
Convolutional Neural Network (CNN)  - image recognitionConvolutional Neural Network (CNN)  - image recognition
Convolutional Neural Network (CNN) - image recognition
YUNG-KUEI CHEN
 
Convolutional Neural Network (CNN)
Convolutional Neural Network (CNN)Convolutional Neural Network (CNN)
Convolutional Neural Network (CNN)
Muhammad Haroon
 
Deep learning - A Visual Introduction
Deep learning - A Visual IntroductionDeep learning - A Visual Introduction
Deep learning - A Visual Introduction
Lukas Masuch
 
Introduction to Recurrent Neural Network
Introduction to Recurrent Neural NetworkIntroduction to Recurrent Neural Network
Introduction to Recurrent Neural Network
Yan Xu
 
CNN Machine learning DeepLearning
CNN Machine learning DeepLearningCNN Machine learning DeepLearning
CNN Machine learning DeepLearning
Abhishek Sharma
 
Recurrent Neural Network (RNN) | RNN LSTM Tutorial | Deep Learning Course | S...
Recurrent Neural Network (RNN) | RNN LSTM Tutorial | Deep Learning Course | S...Recurrent Neural Network (RNN) | RNN LSTM Tutorial | Deep Learning Course | S...
Recurrent Neural Network (RNN) | RNN LSTM Tutorial | Deep Learning Course | S...
Simplilearn
 
Deep learning
Deep learningDeep learning
Deep learning
Ratnakar Pandey
 
Convolutional Neural Network and Its Applications
Convolutional Neural Network and Its ApplicationsConvolutional Neural Network and Its Applications
Convolutional Neural Network and Its Applications
Kasun Chinthaka Piyarathna
 
Back Propagation Neural Network In AI PowerPoint Presentation Slide Templates...
Back Propagation Neural Network In AI PowerPoint Presentation Slide Templates...Back Propagation Neural Network In AI PowerPoint Presentation Slide Templates...
Back Propagation Neural Network In AI PowerPoint Presentation Slide Templates...
SlideTeam
 
Deep Learning Tutorial | Deep Learning Tutorial For Beginners | What Is Deep ...
Deep Learning Tutorial | Deep Learning Tutorial For Beginners | What Is Deep ...Deep Learning Tutorial | Deep Learning Tutorial For Beginners | What Is Deep ...
Deep Learning Tutorial | Deep Learning Tutorial For Beginners | What Is Deep ...
Simplilearn
 
Introduction to Deep Learning
Introduction to Deep LearningIntroduction to Deep Learning
Introduction to Deep Learning
Oswald Campesato
 
Cnn
CnnCnn
Introduction to Neural Networks
Introduction to Neural NetworksIntroduction to Neural Networks
Introduction to Neural Networks
Databricks
 
Support Vector Machines
Support Vector MachinesSupport Vector Machines
Support Vector Machines
CloudxLab
 
Machine Learning - Convolutional Neural Network
Machine Learning - Convolutional Neural NetworkMachine Learning - Convolutional Neural Network
Machine Learning - Convolutional Neural Network
Richard Kuo
 
Multi Layer Network
Multi Layer NetworkMulti Layer Network

What's hot (20)

Convolutional neural network
Convolutional neural network Convolutional neural network
Convolutional neural network
 
Overview of Convolutional Neural Networks
Overview of Convolutional Neural NetworksOverview of Convolutional Neural Networks
Overview of Convolutional Neural Networks
 
CNN and its applications by ketaki
CNN and its applications by ketakiCNN and its applications by ketaki
CNN and its applications by ketaki
 
Convolutional Neural Networks
Convolutional Neural NetworksConvolutional Neural Networks
Convolutional Neural Networks
 
Convolutional Neural Network (CNN) - image recognition
Convolutional Neural Network (CNN)  - image recognitionConvolutional Neural Network (CNN)  - image recognition
Convolutional Neural Network (CNN) - image recognition
 
Convolutional Neural Network (CNN)
Convolutional Neural Network (CNN)Convolutional Neural Network (CNN)
Convolutional Neural Network (CNN)
 
Deep learning - A Visual Introduction
Deep learning - A Visual IntroductionDeep learning - A Visual Introduction
Deep learning - A Visual Introduction
 
Introduction to Recurrent Neural Network
Introduction to Recurrent Neural NetworkIntroduction to Recurrent Neural Network
Introduction to Recurrent Neural Network
 
CNN Machine learning DeepLearning
CNN Machine learning DeepLearningCNN Machine learning DeepLearning
CNN Machine learning DeepLearning
 
Recurrent Neural Network (RNN) | RNN LSTM Tutorial | Deep Learning Course | S...
Recurrent Neural Network (RNN) | RNN LSTM Tutorial | Deep Learning Course | S...Recurrent Neural Network (RNN) | RNN LSTM Tutorial | Deep Learning Course | S...
Recurrent Neural Network (RNN) | RNN LSTM Tutorial | Deep Learning Course | S...
 
Deep learning
Deep learningDeep learning
Deep learning
 
Convolutional Neural Network and Its Applications
Convolutional Neural Network and Its ApplicationsConvolutional Neural Network and Its Applications
Convolutional Neural Network and Its Applications
 
Back Propagation Neural Network In AI PowerPoint Presentation Slide Templates...
Back Propagation Neural Network In AI PowerPoint Presentation Slide Templates...Back Propagation Neural Network In AI PowerPoint Presentation Slide Templates...
Back Propagation Neural Network In AI PowerPoint Presentation Slide Templates...
 
Deep Learning Tutorial | Deep Learning Tutorial For Beginners | What Is Deep ...
Deep Learning Tutorial | Deep Learning Tutorial For Beginners | What Is Deep ...Deep Learning Tutorial | Deep Learning Tutorial For Beginners | What Is Deep ...
Deep Learning Tutorial | Deep Learning Tutorial For Beginners | What Is Deep ...
 
Introduction to Deep Learning
Introduction to Deep LearningIntroduction to Deep Learning
Introduction to Deep Learning
 
Cnn
CnnCnn
Cnn
 
Introduction to Neural Networks
Introduction to Neural NetworksIntroduction to Neural Networks
Introduction to Neural Networks
 
Support Vector Machines
Support Vector MachinesSupport Vector Machines
Support Vector Machines
 
Machine Learning - Convolutional Neural Network
Machine Learning - Convolutional Neural NetworkMachine Learning - Convolutional Neural Network
Machine Learning - Convolutional Neural Network
 
Multi Layer Network
Multi Layer NetworkMulti Layer Network
Multi Layer Network
 

Similar to Visualization of Deep Learning

Deep learning - a primer
Deep learning - a primerDeep learning - a primer
Deep learning - a primer
Uwe Friedrichsen
 
Deep learning - a primer
Deep learning - a primerDeep learning - a primer
Deep learning - a primer
Shirin Elsinghorst
 
Top 10 deep learning algorithms you should know in
Top 10 deep learning algorithms you should know inTop 10 deep learning algorithms you should know in
Top 10 deep learning algorithms you should know in
AmanKumarSingh97
 
Deep learning from a novice perspective
Deep learning from a novice perspectiveDeep learning from a novice perspective
Deep learning from a novice perspective
Anirban Santara
 
Facial Emotion Detection on Children's Emotional Face
Facial Emotion Detection on Children's Emotional FaceFacial Emotion Detection on Children's Emotional Face
Facial Emotion Detection on Children's Emotional Face
Takrim Ul Islam Laskar
 
V2.0 open power ai virtual university deep learning and ai introduction
V2.0 open power ai virtual university   deep learning and ai introductionV2.0 open power ai virtual university   deep learning and ai introduction
V2.0 open power ai virtual university deep learning and ai introduction
Ganesan Narayanasamy
 
Cnn
CnnCnn
introduction to deeplearning
introduction to deeplearningintroduction to deeplearning
introduction to deeplearning
Eyad Alshami
 
Introduction to Deep learning and H2O for beginner's
Introduction to Deep learning and H2O for beginner'sIntroduction to Deep learning and H2O for beginner's
Introduction to Deep learning and H2O for beginner's
Vidyasagar Bhargava
 
Automatic Attendace using convolutional neural network Face Recognition
Automatic Attendace using convolutional neural network Face RecognitionAutomatic Attendace using convolutional neural network Face Recognition
Automatic Attendace using convolutional neural network Face Recognition
vatsal199567
 
MDEC Data Matters Series: machine learning and Deep Learning, A Primer
MDEC Data Matters Series: machine learning and Deep Learning, A PrimerMDEC Data Matters Series: machine learning and Deep Learning, A Primer
MDEC Data Matters Series: machine learning and Deep Learning, A Primer
Poo Kuan Hoong
 
Machine Learning, Deep Learning and Data Analysis Introduction
Machine Learning, Deep Learning and Data Analysis IntroductionMachine Learning, Deep Learning and Data Analysis Introduction
Machine Learning, Deep Learning and Data Analysis Introduction
Te-Yen Liu
 
An introduction to Machine Learning (and a little bit of Deep Learning)
An introduction to Machine Learning (and a little bit of Deep Learning)An introduction to Machine Learning (and a little bit of Deep Learning)
An introduction to Machine Learning (and a little bit of Deep Learning)
Thomas da Silva Paula
 
Development of Deep Learning Architecture
Development of Deep Learning ArchitectureDevelopment of Deep Learning Architecture
Development of Deep Learning Architecture
Pantech ProLabs India Pvt Ltd
 
Introduction to deep learning
Introduction to deep learningIntroduction to deep learning
Introduction to deep learning
Abhishek Bhandwaldar
 
Introduction to deep learning
Introduction to deep learningIntroduction to deep learning
Introduction to deep learning
Amr Rashed
 
Muhammad Usman Akhtar | Ph.D Scholar | Wuhan University | School of Co...
Muhammad Usman Akhtar  |  Ph.D Scholar  |  Wuhan  University  |  School of Co...Muhammad Usman Akhtar  |  Ph.D Scholar  |  Wuhan  University  |  School of Co...
Muhammad Usman Akhtar | Ph.D Scholar | Wuhan University | School of Co...
Wuhan University
 
Separating Hype from Reality in Deep Learning with Sameer Farooqui
 Separating Hype from Reality in Deep Learning with Sameer Farooqui Separating Hype from Reality in Deep Learning with Sameer Farooqui
Separating Hype from Reality in Deep Learning with Sameer Farooqui
Databricks
 
Deep Learning Sample Class (Jon Lederman)
Deep Learning Sample Class (Jon Lederman)Deep Learning Sample Class (Jon Lederman)
Deep Learning Sample Class (Jon Lederman)
Jon Lederman
 
Artificial Neural Network for hand Gesture recognition
Artificial Neural Network for hand Gesture recognitionArtificial Neural Network for hand Gesture recognition
Artificial Neural Network for hand Gesture recognition
Vigneshwer Dhinakaran
 

Similar to Visualization of Deep Learning (20)

Deep learning - a primer
Deep learning - a primerDeep learning - a primer
Deep learning - a primer
 
Deep learning - a primer
Deep learning - a primerDeep learning - a primer
Deep learning - a primer
 
Top 10 deep learning algorithms you should know in
Top 10 deep learning algorithms you should know inTop 10 deep learning algorithms you should know in
Top 10 deep learning algorithms you should know in
 
Deep learning from a novice perspective
Deep learning from a novice perspectiveDeep learning from a novice perspective
Deep learning from a novice perspective
 
Facial Emotion Detection on Children's Emotional Face
Facial Emotion Detection on Children's Emotional FaceFacial Emotion Detection on Children's Emotional Face
Facial Emotion Detection on Children's Emotional Face
 
V2.0 open power ai virtual university deep learning and ai introduction
V2.0 open power ai virtual university   deep learning and ai introductionV2.0 open power ai virtual university   deep learning and ai introduction
V2.0 open power ai virtual university deep learning and ai introduction
 
Cnn
CnnCnn
Cnn
 
introduction to deeplearning
introduction to deeplearningintroduction to deeplearning
introduction to deeplearning
 
Introduction to Deep learning and H2O for beginner's
Introduction to Deep learning and H2O for beginner'sIntroduction to Deep learning and H2O for beginner's
Introduction to Deep learning and H2O for beginner's
 
Automatic Attendace using convolutional neural network Face Recognition
Automatic Attendace using convolutional neural network Face RecognitionAutomatic Attendace using convolutional neural network Face Recognition
Automatic Attendace using convolutional neural network Face Recognition
 
MDEC Data Matters Series: machine learning and Deep Learning, A Primer
MDEC Data Matters Series: machine learning and Deep Learning, A PrimerMDEC Data Matters Series: machine learning and Deep Learning, A Primer
MDEC Data Matters Series: machine learning and Deep Learning, A Primer
 
Machine Learning, Deep Learning and Data Analysis Introduction
Machine Learning, Deep Learning and Data Analysis IntroductionMachine Learning, Deep Learning and Data Analysis Introduction
Machine Learning, Deep Learning and Data Analysis Introduction
 
An introduction to Machine Learning (and a little bit of Deep Learning)
An introduction to Machine Learning (and a little bit of Deep Learning)An introduction to Machine Learning (and a little bit of Deep Learning)
An introduction to Machine Learning (and a little bit of Deep Learning)
 
Development of Deep Learning Architecture
Development of Deep Learning ArchitectureDevelopment of Deep Learning Architecture
Development of Deep Learning Architecture
 
Introduction to deep learning
Introduction to deep learningIntroduction to deep learning
Introduction to deep learning
 
Introduction to deep learning
Introduction to deep learningIntroduction to deep learning
Introduction to deep learning
 
Muhammad Usman Akhtar | Ph.D Scholar | Wuhan University | School of Co...
Muhammad Usman Akhtar  |  Ph.D Scholar  |  Wuhan  University  |  School of Co...Muhammad Usman Akhtar  |  Ph.D Scholar  |  Wuhan  University  |  School of Co...
Muhammad Usman Akhtar | Ph.D Scholar | Wuhan University | School of Co...
 
Separating Hype from Reality in Deep Learning with Sameer Farooqui
 Separating Hype from Reality in Deep Learning with Sameer Farooqui Separating Hype from Reality in Deep Learning with Sameer Farooqui
Separating Hype from Reality in Deep Learning with Sameer Farooqui
 
Deep Learning Sample Class (Jon Lederman)
Deep Learning Sample Class (Jon Lederman)Deep Learning Sample Class (Jon Lederman)
Deep Learning Sample Class (Jon Lederman)
 
Artificial Neural Network for hand Gesture recognition
Artificial Neural Network for hand Gesture recognitionArtificial Neural Network for hand Gesture recognition
Artificial Neural Network for hand Gesture recognition
 

Recently uploaded

Overview on Edible Vaccine: Pros & Cons with Mechanism
Overview on Edible Vaccine: Pros & Cons with MechanismOverview on Edible Vaccine: Pros & Cons with Mechanism
Overview on Edible Vaccine: Pros & Cons with Mechanism
DeeptiGupta154
 
Polish students' mobility in the Czech Republic
Polish students' mobility in the Czech RepublicPolish students' mobility in the Czech Republic
Polish students' mobility in the Czech Republic
Anna Sz.
 
Francesca Gottschalk - How can education support child empowerment.pptx
Francesca Gottschalk - How can education support child empowerment.pptxFrancesca Gottschalk - How can education support child empowerment.pptx
Francesca Gottschalk - How can education support child empowerment.pptx
EduSkills OECD
 
The approach at University of Liverpool.pptx
The approach at University of Liverpool.pptxThe approach at University of Liverpool.pptx
The approach at University of Liverpool.pptx
Jisc
 
Instructions for Submissions thorugh G- Classroom.pptx
Instructions for Submissions thorugh G- Classroom.pptxInstructions for Submissions thorugh G- Classroom.pptx
Instructions for Submissions thorugh G- Classroom.pptx
Jheel Barad
 
678020731-Sumas-y-Restas-Para-Colorear.pdf
678020731-Sumas-y-Restas-Para-Colorear.pdf678020731-Sumas-y-Restas-Para-Colorear.pdf
678020731-Sumas-y-Restas-Para-Colorear.pdf
CarlosHernanMontoyab2
 
Lapbook sobre os Regimes Totalitários.pdf
Lapbook sobre os Regimes Totalitários.pdfLapbook sobre os Regimes Totalitários.pdf
Lapbook sobre os Regimes Totalitários.pdf
Jean Carlos Nunes Paixão
 
Unit 8 - Information and Communication Technology (Paper I).pdf
Unit 8 - Information and Communication Technology (Paper I).pdfUnit 8 - Information and Communication Technology (Paper I).pdf
Unit 8 - Information and Communication Technology (Paper I).pdf
Thiyagu K
 
special B.ed 2nd year old paper_20240531.pdf
special B.ed 2nd year old paper_20240531.pdfspecial B.ed 2nd year old paper_20240531.pdf
special B.ed 2nd year old paper_20240531.pdf
Special education needs
 
The Accursed House by Émile Gaboriau.pptx
The Accursed House by Émile Gaboriau.pptxThe Accursed House by Émile Gaboriau.pptx
The Accursed House by Émile Gaboriau.pptx
DhatriParmar
 
BÀI TẬP BỔ TRỢ TIẾNG ANH GLOBAL SUCCESS LỚP 3 - CẢ NĂM (CÓ FILE NGHE VÀ ĐÁP Á...
BÀI TẬP BỔ TRỢ TIẾNG ANH GLOBAL SUCCESS LỚP 3 - CẢ NĂM (CÓ FILE NGHE VÀ ĐÁP Á...BÀI TẬP BỔ TRỢ TIẾNG ANH GLOBAL SUCCESS LỚP 3 - CẢ NĂM (CÓ FILE NGHE VÀ ĐÁP Á...
BÀI TẬP BỔ TRỢ TIẾNG ANH GLOBAL SUCCESS LỚP 3 - CẢ NĂM (CÓ FILE NGHE VÀ ĐÁP Á...
Nguyen Thanh Tu Collection
 
Introduction to AI for Nonprofits with Tapp Network
Introduction to AI for Nonprofits with Tapp NetworkIntroduction to AI for Nonprofits with Tapp Network
Introduction to AI for Nonprofits with Tapp Network
TechSoup
 
Embracing GenAI - A Strategic Imperative
Embracing GenAI - A Strategic ImperativeEmbracing GenAI - A Strategic Imperative
Embracing GenAI - A Strategic Imperative
Peter Windle
 
Phrasal Verbs.XXXXXXXXXXXXXXXXXXXXXXXXXX
Phrasal Verbs.XXXXXXXXXXXXXXXXXXXXXXXXXXPhrasal Verbs.XXXXXXXXXXXXXXXXXXXXXXXXXX
Phrasal Verbs.XXXXXXXXXXXXXXXXXXXXXXXXXX
MIRIAMSALINAS13
 
2024.06.01 Introducing a competency framework for languag learning materials ...
2024.06.01 Introducing a competency framework for languag learning materials ...2024.06.01 Introducing a competency framework for languag learning materials ...
2024.06.01 Introducing a competency framework for languag learning materials ...
Sandy Millin
 
Model Attribute Check Company Auto Property
Model Attribute  Check Company Auto PropertyModel Attribute  Check Company Auto Property
Model Attribute Check Company Auto Property
Celine George
 
Welcome to TechSoup New Member Orientation and Q&A (May 2024).pdf
Welcome to TechSoup   New Member Orientation and Q&A (May 2024).pdfWelcome to TechSoup   New Member Orientation and Q&A (May 2024).pdf
Welcome to TechSoup New Member Orientation and Q&A (May 2024).pdf
TechSoup
 
The Roman Empire A Historical Colossus.pdf
The Roman Empire A Historical Colossus.pdfThe Roman Empire A Historical Colossus.pdf
The Roman Empire A Historical Colossus.pdf
kaushalkr1407
 
Thesis Statement for students diagnonsed withADHD.ppt
Thesis Statement for students diagnonsed withADHD.pptThesis Statement for students diagnonsed withADHD.ppt
Thesis Statement for students diagnonsed withADHD.ppt
EverAndrsGuerraGuerr
 
"Protectable subject matters, Protection in biotechnology, Protection of othe...
"Protectable subject matters, Protection in biotechnology, Protection of othe..."Protectable subject matters, Protection in biotechnology, Protection of othe...
"Protectable subject matters, Protection in biotechnology, Protection of othe...
SACHIN R KONDAGURI
 

Recently uploaded (20)

Overview on Edible Vaccine: Pros & Cons with Mechanism
Overview on Edible Vaccine: Pros & Cons with MechanismOverview on Edible Vaccine: Pros & Cons with Mechanism
Overview on Edible Vaccine: Pros & Cons with Mechanism
 
Polish students' mobility in the Czech Republic
Polish students' mobility in the Czech RepublicPolish students' mobility in the Czech Republic
Polish students' mobility in the Czech Republic
 
Francesca Gottschalk - How can education support child empowerment.pptx
Francesca Gottschalk - How can education support child empowerment.pptxFrancesca Gottschalk - How can education support child empowerment.pptx
Francesca Gottschalk - How can education support child empowerment.pptx
 
The approach at University of Liverpool.pptx
The approach at University of Liverpool.pptxThe approach at University of Liverpool.pptx
The approach at University of Liverpool.pptx
 
Instructions for Submissions thorugh G- Classroom.pptx
Instructions for Submissions thorugh G- Classroom.pptxInstructions for Submissions thorugh G- Classroom.pptx
Instructions for Submissions thorugh G- Classroom.pptx
 
678020731-Sumas-y-Restas-Para-Colorear.pdf
678020731-Sumas-y-Restas-Para-Colorear.pdf678020731-Sumas-y-Restas-Para-Colorear.pdf
678020731-Sumas-y-Restas-Para-Colorear.pdf
 
Lapbook sobre os Regimes Totalitários.pdf
Lapbook sobre os Regimes Totalitários.pdfLapbook sobre os Regimes Totalitários.pdf
Lapbook sobre os Regimes Totalitários.pdf
 
Unit 8 - Information and Communication Technology (Paper I).pdf
Unit 8 - Information and Communication Technology (Paper I).pdfUnit 8 - Information and Communication Technology (Paper I).pdf
Unit 8 - Information and Communication Technology (Paper I).pdf
 
special B.ed 2nd year old paper_20240531.pdf
special B.ed 2nd year old paper_20240531.pdfspecial B.ed 2nd year old paper_20240531.pdf
special B.ed 2nd year old paper_20240531.pdf
 
The Accursed House by Émile Gaboriau.pptx
The Accursed House by Émile Gaboriau.pptxThe Accursed House by Émile Gaboriau.pptx
The Accursed House by Émile Gaboriau.pptx
 
BÀI TẬP BỔ TRỢ TIẾNG ANH GLOBAL SUCCESS LỚP 3 - CẢ NĂM (CÓ FILE NGHE VÀ ĐÁP Á...
BÀI TẬP BỔ TRỢ TIẾNG ANH GLOBAL SUCCESS LỚP 3 - CẢ NĂM (CÓ FILE NGHE VÀ ĐÁP Á...BÀI TẬP BỔ TRỢ TIẾNG ANH GLOBAL SUCCESS LỚP 3 - CẢ NĂM (CÓ FILE NGHE VÀ ĐÁP Á...
BÀI TẬP BỔ TRỢ TIẾNG ANH GLOBAL SUCCESS LỚP 3 - CẢ NĂM (CÓ FILE NGHE VÀ ĐÁP Á...
 
Introduction to AI for Nonprofits with Tapp Network
Introduction to AI for Nonprofits with Tapp NetworkIntroduction to AI for Nonprofits with Tapp Network
Introduction to AI for Nonprofits with Tapp Network
 
Embracing GenAI - A Strategic Imperative
Embracing GenAI - A Strategic ImperativeEmbracing GenAI - A Strategic Imperative
Embracing GenAI - A Strategic Imperative
 
Phrasal Verbs.XXXXXXXXXXXXXXXXXXXXXXXXXX
Phrasal Verbs.XXXXXXXXXXXXXXXXXXXXXXXXXXPhrasal Verbs.XXXXXXXXXXXXXXXXXXXXXXXXXX
Phrasal Verbs.XXXXXXXXXXXXXXXXXXXXXXXXXX
 
2024.06.01 Introducing a competency framework for languag learning materials ...
2024.06.01 Introducing a competency framework for languag learning materials ...2024.06.01 Introducing a competency framework for languag learning materials ...
2024.06.01 Introducing a competency framework for languag learning materials ...
 
Model Attribute Check Company Auto Property
Model Attribute  Check Company Auto PropertyModel Attribute  Check Company Auto Property
Model Attribute Check Company Auto Property
 
Welcome to TechSoup New Member Orientation and Q&A (May 2024).pdf
Welcome to TechSoup   New Member Orientation and Q&A (May 2024).pdfWelcome to TechSoup   New Member Orientation and Q&A (May 2024).pdf
Welcome to TechSoup New Member Orientation and Q&A (May 2024).pdf
 
The Roman Empire A Historical Colossus.pdf
The Roman Empire A Historical Colossus.pdfThe Roman Empire A Historical Colossus.pdf
The Roman Empire A Historical Colossus.pdf
 
Thesis Statement for students diagnonsed withADHD.ppt
Thesis Statement for students diagnonsed withADHD.pptThesis Statement for students diagnonsed withADHD.ppt
Thesis Statement for students diagnonsed withADHD.ppt
 
"Protectable subject matters, Protection in biotechnology, Protection of othe...
"Protectable subject matters, Protection in biotechnology, Protection of othe..."Protectable subject matters, Protection in biotechnology, Protection of othe...
"Protectable subject matters, Protection in biotechnology, Protection of othe...
 

Visualization of Deep Learning

  • 1. DEEP LEARNING By A. Yamini Saraswathi Y18MTSP401 Under the guidance of Mr. T. Krishna chaithanya (M.Tech)
  • 2. Contents • Artificial intelligence, machine learning • What is deep learning • computer vision features • Fundamentals of deep learning • Back propogation algorithm • Convolutional neural network • Recurrent neural network • Deep learning framework • Applications
  • 3. Artificial intelligence Artficial intelligence is the simulation of human intelligence processes by machines especially computer systems. In this process include learning, reasoning to reach to an approximate or accurate output.
  • 4.
  • 5. Continution.... Fig: Artificial neural network Fig: Biological neuron
  • 6. Continution.... Applications of artficial intelligence • Autonomous vehicles(as drones and self driven cars) • Medical diagnosis • Playing games (such as chess) • Online assistance (such as siri) • Image recognition • Speech recognition • Natural language processing • Etc......
  • 7. Continution... Learning of a neural netwwork can be of two types • Supervised learning:- It is used to perform classification when we want to map input data to output labels. In this data is labeled with classes. • Unsupervised learning:- This learning is used to perform clustering and in this data is not labeled. It is mostly used in exploratory analysis.
  • 8. Machine learning • Which is a subclass of artificial intelligence that automatially learn and improve from the experience without being epxlicitly programmed. • Intelligent systems also built on machine learning algorithms which have the capability to learn from the past experience.
  • 10. Limitations of machine learning • High dimensionality of the data • Unable to solve crucial AI problems • One of the biggest challenge is feature extraciton • Requires only significant amount of data
  • 11. What is deep learning Which is a subclass of machine learning. In this neural network is designed with more than one hidden layer between input and output layer. It reduces the complexity of high dimensions of the data. Fig: Deep neural network
  • 12. Computer vision features • Camera obscura • www.youtube.com/watch?v=10Hay06LJ4 • Black world • SIFT feature • Object recognition • Etc......
  • 13. Fundamentals of deep learning Activation function:- These are really important for artificial neural network to learn and make sense of something really complicated and make non-linear complex functional mapping between inputs and response variable. It should be always non-linear and there are so many types Sigmoid Tanh Softmax Relu Leaky Relu
  • 14. Continution.... Sigmoid It looks like an S- shaped curve. It's mathematical form is f(x) = 1 / 1 + exp(-x) and its range is from 0 to 1 easy to understand and apply but it has some major issues Vanishing gradient It's output is not zero centered Sigmoids saturate and kill gradients Sigmoids have slow convergence
  • 15. Continution.... Tanh It also mostly looked like S-shaped curve and its range is from -1 to 1 and it's mathematical form f(x) = 1-exp(-2x) / 1 + exp(-2x) in this optimizations is easier. it's major issue is Vanishing gradient problem
  • 16. Continution.... Relu It has become more popular in last couple of years. It was recently proved that it had 6 times improvement in convergence compared to Tanh function it also solves vanishing gradient problem there are two major issues • It can be used in hidden layers only • Some gradients may die
  • 17. Continution.... Leaky Relu To fix the problem of dying neurons in the Relu we will use Leaky Relu which is the modification of Relu which introdues a small slope to keep the updates alive
  • 18. Loss function • It is having an important part in ANN which is used to measure the inconsistency between predicted value and actual value. These are very helpful for training the neural network. Y=Predicted output-actual output • Many types of loss functions are present to deal with different types of tasks but most widely used loss function is Mean Square Error(MSE)
  • 19. Optimization Weights and bias are modifiedusing a function calledoptimization function in such a way error is minimized. For updating weights we use an equation Optimazation algorithms are pf 2 types • First order optmization algorithm • Second order optimization algorithm
  • 20. Optimzation algorithms Gradient descent is the most popularly used optmization algorithm. Which is of two types • Stochastich gradient descent θ=θ−η∇J(θ) where ∇J(θ) is the Gradient of Loss function J(θ) w.r.t parameters 'θ’ • Mini batch gradient descent
  • 21. Weight initialization The method that we follow for weight initilalization is 'Xavier intialization'. This method tries to intialize the weights with a smarter value
  • 22. Normalization It is most important part in ANN because because most of the data obtained from the real world are very distant each other. So to keep the data at a comparable range we need to normalize the input data of input layer. Common normalization approach includes. • Statistical normalization • Sigmoid normalization
  • 23. Batch normalizaton • Training deep neural network with more number of layers is more challenging one to avoid this problem 'Batch normalization' is introduced. • It trains deep neural network that standardizes the inputs to a layernfor each mini-batch and Benifits of this batch normalization are • Network train • Makes weights easier to initialize • Simplifies the creation of deeper network • Makes more activation functions viable
  • 25. Regularization When we build a neural network it performs well for trained data but not on the test data. This is the sign of over fitting to reduce this problem there are two ways • Use more data • Regularization Getting more data is sometimes impossible and expoensive so for this purpose we use 'Regularization' So to solve this we should make our model generalize over the trained data which is done using various regularization techniques.
  • 26. Continuation... Regularization techniques • L1 & L2 regularization Which updates general cost function by adding another term known as regularization term There is a small difference between L1 and L2 in the equation. Mostly L1 is preferred
  • 27. Continuation.... • Dropout regularization This is mostly used one in the deep learning. At every iteration it randomly selects some nodes and removes remaining with all of their incoming signals and outgoing connections So at each iteration has a different set of nodes which result a different set of output At last we ensemble models which usually perform better than a normal neural network Fig: Drop out Neural network
  • 28. Back propogation algorithm Neural network is trained by 'Back propogation algorithm' after forwarding inputs to neural network then inputs and weights are summed and multiplied then apply to activation function at last we predict the error so to reduce this error we need to update weights using gradient descent and bias which propogate backwwards to network
  • 29. Convolutional neural network In todays, convnets are used in every where it has been very effective in such as image classification and image recognition it has been very successful in identifying faces, objects and traffic signs LeNet was one of the first convolutional neural network which helped the path to deep learning Convent is mainly build based on 3 layers • Convolution layer & Non linearity • Pooling layer • Fully connected layer
  • 31. Continution.... Convolution operation Fig:Input image Fig: Kernal filter Fig: Feature map K=1
  • 32. Continution.... Padding & pooling operation Fig: Pooling operation Fig: Padding operation
  • 35. Recurrent Neural Network • RNN are also a feed forward neural network, however with ecurrent meomory loops which takes input from the previous and /or same layers of states • This gives them a unique capability to model along the time dimension and arbitary sequence of even inputs. • RNN's are used for sequenced data analysis such as time series, sentimental nalysis, NLP, language translation, speech recognition, image captioning, script recognition and other othings. • These are aslo calles networks with memory as the previous inputs or states may persist in the model to do a sequential analysis. These memories become input as well
  • 36. Google Net • It is almost different compared to other architectures like Alex Net • It is the winner of Image Net 2014 and can train network faster which is developed by Google • In this global average pooling is used at the end of architecture it leads to reduce number of weights are reduced and accuracy increases to 0.6%
  • 38. Deep Learning Framework • Deep learning frameworks were built in academia in first generation Caffe from Berkley university Torch from NYU and also collaboration with Facebook • In second generation all deep learning frameworks were developed from industry's Caffe2 from Facebook Pytarch from Facebook Tensor flow is from Facebook MXNet from Amazon CNTK form Microsoft
  • 39. Importance of Deep learning frame work • Easily can build big computational graphs • Easily compute gradients in computational graphs • Run it all efficiently n GPU
  • 40. Continuation... • This framework is an open source which aims to simplify the implementation of compels and large scale deep learning models using these amazing frameworks we can build complex models like CNN. Some types of frameworks listed below. Tensor flow Keras Pytorch Caffe Deeplearning4j • One of the most popular frame work is 'Tensorflow'
  • 41. Applications • Automatic speech recognition • Image recognition • Visual art processing • Natural language processing • Drug discovery and toxicology • Bioinformatics • Medical Image Analysis • Mobile advertising • Image restoration • Financial fraud detection • Military
  • 42. Google Arts & Culture app This app uses automatic image analysis to compare lot of pictures in the museums and in institutions throughout the world with your picture. For this process mainly deep learning is mainly involved and in the next slide I have taken the pictures practically.