Introduction of Deep Learning

Lecture: 01
 Introduction of deep learning
Applications of Deep Learning
Why we use Deep Learning?
Characteristics of Deep Learning
Positive of Deep Learning

Agenda Deep Learning Technique

Working of Deep Learning
Deep Learning Technique
How to Create Deep Learning Models?
Perceptron Learning Algorithm
Perceptron Algorithm Block Diagram
  Deep Learning is a part of machine learning, which is a subset of
Artificial Intelligence. It enables us to extract the information from the
layers present in its architecture. It is used in Image Recognition, Fraud
Introduction of Detection, News Analysis, Stock Analysis, Self-driving cars, Healthcare
like cancer image analysis, etc. By inputting more data in the network
Deep Learning the layers get trained very well. They can be classified into Supervised,
Semi-Supervised and Unsupervised categories. Each layer is known for
extracting information specifically.
 example in Image recognition, the first layer will find the edge, lines,
etc, second layer like eye, ear, nose, etc.
APPLICATIONS OF
DEEP LEARNING
  Healthcare
 From Medical image analysis to curing diseases, Deep Learning played a
huge role especially when GPU-processors are present. It also helps
APPLICATIONS OF Physicians, Clinicians, and doctors to help the patients out of danger, and also
DEEP LEARNING they can diagnose and treat the patients with apt medicines.
  Stock Analysis
 Quantitative Equity Analysts are getting more benefits especially to find
the trends for a particular stock whether it will be bullish or bearish and
they can use many more factors like no of transactions made, no of
APPLICATIONS OF buyers, no of sellers, previous day closing balance, etc when training
DEEP LEARNING the deep learning layers. Qualitative Equity Analysts use factors like
return on equity, P/E ratio, Return on Asset, Dividend, Return on
Capital Employed, Profit per Employee, Total Cash, etc when training
the deep learning layers.
  Fraud Detection
 These days, hackers especially those based out of the dark- web have
found ways to steal money digitally across the globe using different
software. Deep learning will learn to find these types of fraudulent
APPLICATIONS OF transactions in the web using a lot of factors like Router information, IP
DEEP LEARNING addresses, etc. Autoencoders also help financial institutions saving
billions of dollars in terms of cost. These types of fraudulent
transactions can also be detected by finding the outliers and
investigating the same.
  Image Recognition
 Suppose say the city police department has a people database of the city
and they wanted to know in public gatherings like who is involved in
APPLICATIONS OF the crimes, violence using public webcam available in streets this deep
DEEP LEARNING learning using CNN (Convolution Neural networks) helps a lot in
finding the person who was involved in the act.
  News Analysis
 These days the government takes a lot of effort especially in controlling the
spread of fake news and origin of it. Also during poll surveys like who would
win elections in terms of popularity, which candidate been shared by most
APPLICATIONS OF people in social media etc and analysis of tweets made by country people
DEEP LEARNING using all these variables we can predict the outcomes in deep learning, but
also there are some limitations to it, we don’t know the data authenticity
whether its genuine or fake. etc or whether the necessary information been
spread by bots.
  Self Driving Cars
 Self-driving cars use Deep Learning by analyzing the data captured in the cars
made in different terrains like mountains, deserts, Land, etc. Data can be
APPLICATIONS OF captured from sensors, public cams, etc which will be helpful in testing and
DEEP LEARNING implementation of self-driving cars. The system must able to ensure all the
scenarios been handled well in training.
  To help improve the efficiency of predictions, to find the best possible
Why we use Deep outcomes and for model optimization. When the data is huge, to reduce the
Learning? cost in the company in terms of insurance, sales, profit, etc. Deep learning
can be very useful when there is no particular structure to data means to
analyze data from audio, video, image, numbers, document processing,
etc.
  Supervised, Semi-Supervised or Unsupervised
 When the category labels are present while you train the data
then it is Supervised learning. Algorithms like Linear regression. Logistic
Characteristics of regression, decision trees use Supervised Learning. When category labels are
not known while you train data then it is unsupervised learning. Algorithms
Deep Learning like Cluster Analysis, K means clustering, Anomaly detection uses
Unsupervised Learning. The data set consists of both labeled and unlabelled
data then we call it is Semi-Supervised learning. Graph-based models,
Generative models, cluster assumptions, continuity assumptions use Semi-
Supervised learning.
  Huge Amount of Resources
 It needs advanced Graphical Processing Units for processing heavy
Characteristics of Deep workloads. A huge amount of data needs to be processed like Big data in the
Learning form of structured or unstructured data. Sometimes more time is also required
to process the data, it depends on the amount of data fed in.
  Large Amount of Layers in Model

Characteristics of  A huge amount of layers like input, activation, the output will be
required, sometimes the output of one layer can be input to another
Deep Learning layer by making few small findings and then these findings are
summed up finally in the softmax layer to find out a broader
classification for final output.
  Optimizing Hyper-parameters
 Hyperparameters like no of epochs, Batch size, No of layers,
Characteristics of Learning rate, needs to be tuned well for successful Model accuracy
Deep Learning because it creates a link between layer predictions to final output
prediction. Over-fitting and under-fitting can be well handled with
hyper-parameters.
  Cost Function
 It says how well the model performance in prediction and accuracy.
Characteristics of For each iteration in Deep Learning Model, the goal is to minimize
the cost when compared to previous iterations. Mean absolute error,
Deep Learning Mean Squared Error, Hinge loss, Cross entropy are different types
according to different algorithms used.
 • Solve Complex problems like Audio processing in Amazon echo,
Image recognition, etc, reduce the need for feature extraction,
automated tasks wherein predictions can be done in less time using
Keras and Tensorflow.
Positive Aspects of • Parallel computing can be done thus reducing overheads.
Deep Learning • Models can be trained on a huge amount of data and the model gets
better with more data.
• High-Quality Predictions when compared with humans by training
tirelessly.
• Works well-unstructured data like video clips, documents, sensor data,
webcam data, etc
  Deep Learning Techniques are the techniques used for mimicking the
functionality of human brain, by creating models that are used in
Deep Learning classifications from text, images and sounds. These models are made up
Technique of several layers of hidden layer also know as Neural network which
can extract features from the data, each layer of these neural networks
starting from the left-most layer to the rightmost layer extract a low-
level feature like edge and subsequently make predictions accurately.
  Deep Learning methods use Neural Networks. So, they are often
Working of Deep referred to as Deep Neural Networks. Deep or hidden Neural Networks
have multiple hidden layers of deep networks. Deep Learning trains the
Learning AI to predict output with the help of certain inputs or hidden network
layers. These networks are trained by large labeled datasets and learn
features from the data itself. Both
Supervised and Unsupervised Learning works in training the data and
generating features.
Working of Deep
Learning

• The above circles are neurons that are

interconnected.
• There are 3 types of neurons:
  Input layer
 The input layer of a neural network is composed of artificial input neurons,
Working of Deep and brings the

Learning initial data into the system for further processing by subsequent layers of
artificial neurons.
 Hidden layer(s)
 Hidden layer(s) are the secret sauce of your network. They allow you to model
complex data thanks to their nodes/neurons. They are “hidden” because the
true values of their nodes are unknown in the training dataset. In fact, we only
know the input and output. Each neural network has at least one hidden layer.
  Output layer
 The output layer is the final layer in the neural network where desired
predictions are obtained. There is one output layer in a neural network
Working of Deep that produces the desired final prediction. It has its own set of weights
Learning and biases that are applied before the final output is derived
 The input layer gets the input data and passes the input to the first
hidden layer. The mathematical calculations are performed on the
inputdata. Finally, the output layer gives the findings.
  Deep Learning algorithms run through several layers of the hidden
layer(s) or Neural Networks. So, they learn deeply about the images
for accurate prediction. Every layer learns and detects low-level
features like edges and subsequently, the new layer merges with the
features of the earlier layer for better representation. For example, a
Deep Learning middle layer might detect any edge of the object while the hidden
Technique layer will detect the full object or image.
 This technique is efficient with large and complex data. If the data
is small or incomplete, DL becomes incapable to work with new
data.
 There are some Deep Learning Networks as follows:
 • Unsupervised Pre-trained Network: It is a basic model with 3
layers: input, hidden and output layer. The network is trained to
reconstruct the input and then hidden layers learn from the inputs to
Deep Learning gather information and finally, features are extracted from the
image.
Technique
• Conventional Neural Network: As standard Neural Network, it
has a convolution inside for edge detection and accurate recognition
of objects.
 • Recurrent Neural Network: In this technique, the output from the
previous stage is used as input for the next or current stage. RNN stores
the information in context nodes to learn the input data and produce the
output. For example, to complete a sentence we need words. i.e to
Deep Learning predict the next word, previous words are required which needs to be
Technique remembered. RNN basically solves this type of problem.
• Recursive Neural Networks: It is a hierarchical model where the input
is a tree-like structure. This kind of network is created by applying the
same set of weights over the assembly of inputs.
  Deep Learning algorithms are made by connecting layers between
them. The first step above is the input layer followed by the hidden
layer(s) and the output layer. Each layer is composed of interlinked
neurons. The network consumes a large amount of input data to operate
them through multiple layers.
How to Create  To create a Deep Learning model, the following steps are needed:
Deep Learning • Understanding the problem
Models? • Identify data
• Select the algorithm
• Train the model
• Test the model
  Perceptron Algorithm is used in a supervised machine learning domain
for classification. In classification, there are two types of linear
Introduction to Perceptron classification and no-linear classification. Linear classification is
Learning Algorithm nothing but if we can classify the data set by drawing a simple straight
line then it can be called a linear binary classifier. Whereas if we cannot
classify the data set by drawing a simple straight line then it can be
called a non-linear binary classifier.
Perceptron Algorithm
Block Diagram
  Input: All the features of the model we want to train the neural network
will be passed as the input to it, Like the set of features [X1, X2,
X3…..Xn]. Where n represents the total number of features and X
represents the value of the feature.
Perceptron  Weights: Initially, we have to pass some random values as values to the
Algorithm Block weights and these values get automatically updated after each training
error that is the values are generated during the training of the model. In
Diagram some cases, weights can also be called as weight coefficients.
Explanation  Weights Sum: Each input value will be first multiplied with the weight
assigned to it and the sum of all the multiplied values is known as a
weighted sum.
 Weights sum = ∑Wi * Xi (from i=1 to i=n) + (W0 * 1)
  https://www.educba.com/deep-learning/
 https://www.educba.com/deep-learning-toolbox/
 https://www.educba.com/deep-learning-software
Supporting Material  https://towardsdatascience.com
 https://www.edureka.com
 https://www.geeksforgeeks.org
Q/A Thanks