MODULE - 1 Basics of AI, ML & DL

STRUCTURE
1.1 Introduction to Artificial Intelligence
1.1.1 Applications of Artificial Intelligence
1.1.2 History of Artificial Intelligence
1.1.3 Types of Artificial Intelligence
1.2 Introduction to Machine Learning
1.2.1 Techniques in Machine Learning
1.2.2 Applications of Machine Learning
1.2.3 Machine Learning Algorithms
1.3 Introduction to Deep Learning
1.4 Similarities and Differences Between AI, ML & DL
1.5 Summary
1.6 Keywords
1.7 Self-Assessment Questions

LEARNING OUTCOMES
1. Understand the basics of Artificial Intelligence
2. Identify the goals with advantages and disadvantages of Artificial Intelligence
3. Understand the basics of Machine Learning
4. Interpret techniques used in Machine Learning
5. Understand the basics of Machine Learning
6. Identify similarities in AI, ML & DL
7. Differentiate Between AI, ML & DL

1
1.1 INTRODUCTION TO ARTIFICIAL INTELLIGENCE
Technology is rapidly evolving in today's world, and we are constantly in contact with new
technologies. Here, one of the booming computer science technologies is Artificial
Intelligence, which is projected to herald a fresh revolution in the world by creating
intelligent machines. Artificial intelligence is now all around us. It is currently working on
a variety of subfields ranging from general to specific, such as self-driving cars, chess,
proving theorems, music, painting, and so on. AI is one of the most fascinating and universal
fields of computer science, with a bright future ahead of it. AI has the potential to make a
machine work as if it were a human.
Artificial intelligence is made up of the words "artificial" and "intelligence," where
"artificial" refers to something that is "man-made" and "intelligent" refers to something that
has "thinking ability." We can therefore describe AI as "a branch of computer science by
which we can build intelligent machines that can behave like a human, think like humans,
and be capable of making decisions."
Artificial intelligence is the ability of a computer to mimic human abilities like learning,
reasoning, and problem-solving. The awesomeness of artificial intelligence is that it
eliminates the need to pre-program a machine to perform a task. However, you can still
build a machine with pre-programmed algorithms that can perform tasks using their own
intelligence.
It is believed that artificial intelligence (AI) is not a latest invention, and some people claim
that mechanical men who could function and act like humans existed in ancient times in
accordance with Greek myth.
Before learning about artificial intelligence, we should be aware of its significance and the
reasons for learning about it. Here are some primary justifications for learning about AI:
• With the aid of AI, it is possible to develop tools or software that can quickly and
accurately address a variety of real-world issues, including transportation,
marketing, and health-related ones.
• Cortana, Siri, Google Assistant, and other personal virtual assistants are all created
with the aid of AI.
• With the aid of AI, it is possible to create robots that can function in settings where
human life may be in jeopardy.
• AI paves the way for new gadgets, technologies, and opportunities.

The main goals of Artificial Intelligence:

 Simulate human intelligence
 Complete tasks that require a high level of knowledge.
 An intelligent link between perception and action
 Creating a machine capable of performing tasks that require human intelligence,
such as:
2
 • demonstrating a theorem
• Enjoying a game of chess
• Consider a surgical procedure.
• Driving a car in heavy traffic
 Developing a system capable of intelligent behaviour, self-learning, demonstrating,
explaining, and advising its user.

Even though it is so vast and needs many different elements, artificial intelligence is not
solely a branch of computer science. Intelligence is an intangible component of our brain
that is a combination of Reasoning, learning, problem-solving perception, language
comprehension, etc. In order to create artificial intelligence, we must first understand how
intelligence is made.
Artificial intelligence needs knowledge from a wide range of disciplines, including:
mathematics, biology, psychology, sociology, computer science, neuron research, and
statistics, in order to achieve the aforementioned goals for a machine or piece of software.

The primary benefits of artificial intelligence are:

 High Accuracy with Less Errors: AI machines or systems tend to make decisions
with a high degree of accuracy based on prior knowledge or experience.
 High-Speed: Because AI systems are capable of making decisions quickly and at a
high rate of speed, they are capable of defeating chess champions.
 Exceptional accuracy: AI systems are incredibly accurate and capable of repeating
the same action many times.
 Helpful in risky situations: AI machines can be useful in circumstances where
using humans would be risky, such as defusing bombs and exploring the ocean
bottom.
 Digital Assistant: AI can be very helpful in providing users with a digital assistant.
For example, many E-commerce websites already use AI technology to display
goods based on customer preferences.
 Useful as a public utility: AI can be very helpful for services that benefit the general
public, such as self-driving cars that can make travel safer and easier, facial
recognition for security, natural language processing to converse with people in their
own tongue, etc.
Every technology has drawbacks, and artificial intelligence is no exception. Despite being
such a beneficial technology, it does have some drawbacks that we must keep in mind when
developing an AI system. AI's drawbacks include:
 Expensive: AI hardware and software are very expensive, as they require extensive
maintenance to meet current world requirements.
3
  Can't think outside the box: Even though AI is making smarter machines, they still
can't work outside the box because the robot will only do the work for which they
have been trained or programmed.
 No feelings or emotions: AI machines can be excellent performers, but because
they lack feelings, they cannot form emotional attachments with humans and can
sometimes be harmful to users if proper precautions are not taken.
 Increased reliance on machines: As technology advances, people become more
reliant on devices, and as a result, their mental capabilities deteriorate.
 No Original Creativity: Although humans are very creative and can imagine new
ideas, AI machines cannot compete with human intelligence and cannot be creative
or imaginative.

1.1.1 Applications of Artificial Intelligence

In today's culture, artificial intelligence is used in a variety of ways. Because it can
effectively solve complex problems in a variety of sectors, including healthcare,
entertainment, finance, and education, it is becoming increasingly important in the modern
world. AI is enhancing the comfort and speed of our everyday lives.
Some industries that use artificial intelligence include astronomy, healthcare, gaming,
finance, data security, social media, travel & transportation, the automotive industry, and
robotics.
 Astronomy
In order to tackle difficult universe problems, artificial intelligence can be very helpful.
Understanding the world, including its workings and origin, can be aided by AI technology.
 Healthcare
In the last five to ten years, AI has become more beneficial for the healthcare sector and is
expected to have a big effect on this sector.
AI is being used in the healthcare sector to diagnose patients more quickly and accurately
than people. AI can assist physicians with diagnoses and can alert them when patients'
conditions deteriorate so that treatment can be administered before the patient is
hospitalized.
 Gaming
AI can be utilized in games. The AI systems are capable of playing strategic games like
chess, which require a lot of creative thinking on the part of the machine.
 Finance
The finance and AI sectors make the best partners. Automation, chatbots, adaptive
intelligence, algorithm trading, and machine learning are all being applied to financial
operations in the finance sector.

4
  Data Security
Every business must prioritize data protection, and in the digital age, cyberattacks are
increasing significantly. Your data can be made more private and safe with the help of AI.
Examples like the AEG assistant and the AI2 Platform are used to more accurately identify
software bugs and cyber-attacks.
 Social Media
There are billions of user profiles on social media platforms like Facebook, Twitter, and
Snapchat, all of which need to be stored and handled very effectively. Massive quantities of
data can be managed and organized by AI. A lot of data can be analyzed by AI to find the
newest hashtags, patterns, and user requirements.
 Travel & Transport
The demand for AI in the tourism industry is growing rapidly. AI is capable of performing
a variety of travel-related tasks, including planning trips and recommending hotels, flights,
and the best paths to clients. AI-powered chatbots are being used in the travel industry to
engage with customers in a human-like manner for better and quicker responses.
 Automotive Industry
Some industries in the automotive sector are utilizing AI to give their users virtual assistants
for improved performance. Tesla, for example, recently unveiled TeslaBot, a clever virtual
helper.
Various industries are presently working to develop self-driving cars that can increase the
safety and security of your journey.
 Robotics:
In robotics, artificial intelligence plays a remarkable part. The majority of the time, general-
purpose robots are designed to carry out some repetitive tasks, but with the aid of AI, we
can develop intelligent robots that can carry out tasks based on their own experiences rather
than being pre-programmed.
Robots that resemble humans, such as Erica and Sophia, are the finest examples of artificial
intelligence in robotics. These robots can speak and act in human-like ways.
 Entertainment
Our everyday lives now include some AI-based applications that work with entertainment
services like Netflix or Amazon. These services display suggestions for programs or shows
using ML/AI algorithms.
 Agriculture
For the best results, agriculture needs a variety of resources, labor, funds, and time. Today's
agribusiness is increasingly digital, and AI is becoming more prevalent in this industry. AI
is being used in farmland for robotic farming, crop monitoring, and predictive analysis. AI
in farming can be very beneficial to producers.
 E-commerce
5
The e-commerce sector now has a competitive advantage due to AI, which is also becoming
more valuable to the sector. AI is assisting customers in finding related goods with
suggested size, color, or brand.
 Education:
Grading can be automated by AI, giving the tutor more time to educate. A teaching assistant
AI chatbot can interact with pupils.
In the future, AI might serve as a convenient personal virtual tutor for pupils, available at
anytime and anywhere.
1.1.2 History of Artificial Intelligence
For researchers, the term "artificial intelligence" and the related technology are not novel.
As opposed to what you might think, this technology is much ancient. Even in Greek and
Egyptian myths, there are stories about mechanical beings. Let's talk about some key
moments in AI history that illustrate its growth from its inception to the present day.

Artificial intelligence's development between 1943 and 1952:

 Warren McCulloch and Walter Pits completed the first piece of work that is now
known as AI in 1943. They put forth a concept of synthetic neurons.

 In 1949, Donald Hebb showed a rule for updating the strength of the connections
between neurons. Hebbian learning is the modern name for his law.

 In the year 1950: English scientist Alan Turing invented machine learning in that
year. In his book "Computing Machinery and Intelligence," Alan Turing outlined a
test. A Turing test can be used to determine whether a machine is capable of
behaving intelligently on par with a person.

Between 1952 and 1956, artificial intelligence began to emerge:

 In the year 1955, Allen Newell and Herbert A. Simon developed "Logic Theorist,"
the first artificial intelligence software. In addition to finding novel and better proofs
for some theorems, this program had proven 38 of 52 mathematical theorems.

 John McCarthy, an American computer scientist, coined the term "artificial

intelligence" in 1956 at the Dartmouth Conference. AI was first recognized as a
legitimate scholarly discipline.

 High-level programming languages like FORTRAN, LISP, and COBOL were

created at that period. And at that time, interest in AI was at an all-time peak.

The era of prosperity- Early zeal between 1956 and 1974:

6
  In 1966, academics focused on creating algorithms that could resolve mathematical
puzzles. o In 1966, Joseph Weizenbaum invented the first robot, which he called
ELIZA.

 The first sentient humanoid robot, known as WABOT-1, was created in Japan in
1972.

The first AI cold period ran from the years 1974 to 1980:
 The term "AI winter" describes an era of time when computer scientists struggled
with a severe lack of government funding for AI research.
 Public interest in artificial intelligence dropped during AI winters.

AI experienced a surge from 1980 to 1987:

After a long absence, AI returned with "Expert System" in the year 1980. Expert systems
that can make decisions like a person expert have been programmed.

The American Association of Artificial Intelligence conducted its first national conference
at Stanford University in the year 1980.

During the second AI Winter from 1987 to 1993:

The second AI Winter period spanned the years 1987 to 1993.

Again, due to high costs and ineffective results, investors and the government ceased
funding AI research. A very cost-effective expert system was XCON.

The development of intelligent entities between 1993 and 2011:

 In 1997, IBM Deep Blue defeated Gary Kasparov, the reigning world chess
champion, becoming the first machine to do so.

 The Roomba vacuum cleaner marked the first time artificial intelligence (AI) made
its way into a household in the year 2002.

 Beginning in the year 2006, AI entered the business sector. Additionally, businesses
like Facebook, Twitter, and Netflix began utilizing AI.

Big data, artificial general intelligence, and deep learning from 2011 to the present
 In 2011, IBM's Watson won jeopardy, a game show where contestants had to answer
challenging questions and riddles. Watson had demonstrated its ability to
comprehend natural language and rapidly find answers to challenging problems.
 In the year 2012, Google introduced "Google Now," a function for Android apps
that could predict information for users.
 In 2014: In 2014, the chatbot "Eugene Goostman" took first place in the infamous
"Turing test" challenge.

7
  In the year 2018: The IBM "Project Debater" excelled in a discussion on difficult
subjects with two master debaters.

 Google showed off the "Duplex" AI program, a virtual assistant that had accepted a
hairdresser appointment on call while the person on the other end of the line was
speaking with a human.

The degree of AI development today is astounding. Deep learning, big data, and data
science are currently in vogue like never before. These days, businesses like Google,
Facebook, IBM, and Amazon use AI to develop incredible technology. Artificial
intelligence has a bright future and will be highly intelligent.

1.1.3 Types of Artificial Intelligence

While there are many different ways to categorize artificial intelligence, the two major
classifications are based on the capabilities and the functionalities of the technology.

AI type-1: Based on Capabilities:

1. Weak AI or Narrow AI:

 Narrow AI is a subset of AI that is intelligent enough to carry out specific tasks. In
the field of artificial intelligence, narrow AI is the most prevalent and readily
accessible AI.
 Because it is only trained to perform one specific job, narrow AI cannot perform
beyond its field or limitations. As a result, it is also known as weak AI. If narrow
AI exceeds its bounds, it may falter in unexpected ways.
 Apple Siri is an excellent example of narrow artificial intelligence, but it only
performs a small set of pre-defined tasks.
 IBM's Watson supercomputer also falls under the category of narrow AI because it
employs a combination of machine learning, natural language processing, and
expert systems.
 Examples of Narrow AI include are playing chess, purchasing suggestions on e-
commerce sites, self-driving cars, speech recognition, and image recognition.
2. General AI:
 A type of intelligence known as general artificial intelligence is capable of handling
any intellectual job as effectively as a human.
 The goal of general AI is to create a system that is intelligent enough to reason like
a human on its own.
 It's a good idea to have a backup plan in place, especially if you're planning to travel.
 The goal of current global study is to create machines with general artificial
intelligence.
 Because general AI systems are still in the research stage, it will take a long period
and a lot of work to develop them.
8
3. Super AI:
 Super AI is a level of system intelligence where computers are capable of
outperforming humans at any job by combining cognitive abilities. It results from
general AI.
 Strong AI should be able to think, reason, solve puzzles, make decisions, plan, learn,
and communicate on its own, among other important traits.
 Super AI is still just an idealized idea in the world of artificial intelligence. Real-
world implementation of such networks is still a challenging endeavor.

AI type-2: Based on functionality:

1. Reactive Machines
 The most fundamental varieties of artificial intelligence are purely reactive systems.
 These AI systems don't keep track of memories or previous encounters for future
use.
 These machines only concentrate on present situations and respond in the best way
they can.
 An illustration of a reactive machine is IBM's Deep Blue system.
 AlphaGo from Google is another illustration of a reactive machine.
2. Limited Memory
 Machines with a small amount of memory can temporarily hold data or memories.
 These devices can only access stored data for a brief amount of time.
 One of the finest examples of Limited Memory systems is self-driving cars. These
vehicles can store information for road navigation, including the speed limit,
distance to other vehicles, and recent speeds of those close.
3. Theory of Mind
 Theories of Mind AI should be able to engage socially with humans and
comprehend human emotions, people, and beliefs.

 Although this kind of AI devices have not yet been created, researchers are working
hard to advance their development.

4. Self-Awareness
 Self-awareness AI is the future of Artificial Intelligence. They will be extremely
intelligent and possess cognition, feelings, and self-awareness.
 These devices will be more intelligent than the human intellect.
 Self-Awareness AI is still a theoretical idea and does not currently exist in actuality.

9
1.2 INTRODUCTION TO MACHINE LEARNING

Machine learning is a field of technology that teaches computers to carry out a variety of
tasks, including predictions, suggestions, estimations, etc., based on previous experience or
historical data. Through the use of predicted data and prior experience, machine learning
trains computers to act like people.

Task, Experience, and Performance are the three main facets of machine learning.

 The main problem in which we are concerned is defined as a task. This task/problem
may be linked to predictions, recommendations, and estimations, among other
things.

 Experience is described as learning from historical or previous data and applying

that knowledge to estimate and resolve future tasks.

 Performance is defined as a machine's ability to solve any machine-learning task or

issue and provide the best possible result. However, performance is affected by the
nature of the machine learning issues.

1.2.1 Techniques in Machine Learning

Machine Learning techniques are broadly classified into four groups:

1. Supervised Learning
 When a machine has sample data, that is, input and output data with accurate labels,
supervised learning is practical. Using some labels and tags, accurate labels are used
to verify the model's accuracy. Using labeled examples from the past and supervised
learning techniques, we can forecast future occurrences. It begins by analyzing the
known training dataset and then adds an inferred function that forecasts the values
of the output. Additionally, it anticipates mistakes throughout the entire learning
process and uses algorithms to rectify them.
 Assume we have a collection of pictures with the title "dog" on them. These dog
images are used to train a machine learning algorithm so that it can differentiate
between them whether an image is a dog or not.
2. Unsupervised Learning
 In unsupervised learning, a machine learns using only a limited set of input samples
or labels, with no knowledge of the output. In contrast to supervised learning, a
computer may not always produce the right results because the training data is
neither classified nor labeled.
 Although less common in real-world business settings, unsupervised learning aids
in data exploration and can be used to deduce hidden structures from unlabeled data.
 Example: Assume for the sake of illustration that a machine has been trained on a
collection of documents with various categories (Type A, B, and C), and we need
10
 to group them appropriately. The machine can classify these datasets into types A,
B, and C because it is only given input samples or no output, so it is not crucial
whether the classification is accurate or not.
3. Reinforcement Learning
 A feedback-based machine learning method is reinforcement learning. In this kind
of learning, agents (computer programs) must investigate their surroundings, take
actions, and then receive rewards as feedback for their actions. They receive a
positive reward for every good deed and a negative benefit for every bad deed. A
reinforcement learning agent's objective is to optimize the good outcomes. The
agent can only learn from experience because there is no annotated data.
4. Semi-supervised Learning
 Semi-supervised learning is a middle ground between supervised and autonomous
learning techniques. It conducts actions on datasets having few labels as well as
unlabeled data. However, the material is typically unlabeled. Because labels are
expensive but may not be necessary for business purposes, it also lowers the cost of
the machine learning model. It also improves the machine learning model's
efficiency and accuracy.
 Data scientists can surmount the shortcomings of supervised and unsupervised
learning with the aid of semi-supervised learning. Some significant uses of semi-
supervised learning include speech analysis, web content classification, protein
sequence classification, text document classifiers, etc.

1.2.2 Applications of Machine Learning

Almost every industry, including those in healthcare, marketing, banking, infrastructure,
automation, etc., uses machine learning extensively. Some significant machine learning
examples from the actual world include:

• Healthcare and Medical Diagnosis

• Marketing
• Self-driving cars
• Speech Recognition
• Traffic Prediction
• Image Recognition
• Product Recommendations
• Automatic Translation
• Virtual Assistant
• Email Spam and Malware Filtering

1.2.3 Machine Learning Algorithms

Several frequently used machine learning algorithms include linear regression, logistic
regression, K Nearest Neighbor (KNN), K-Means Clustering, decision trees, random
11
forests, support vector machines (SVM), and naive bayes.

 Linear Regression:

One of the most straightforward and well-liked machine learning algorithms that a data
analyst suggests is linear regression. Making predictions for actual variables like expertise,
salary, cost, etc. is used for predictive analysis.

Linear regression is a statistical method that depicts the linear connection between two or
more variables, either dependent or independent. The slope of this graph is referred to as
the Line of Regression, and it demonstrates how the dependent variable's value varies in
relation to the independent variable.

Fig. Linear regression

In Linear Regression,

y= a0+a1x+ ε

Y= Dependent Variable

X= Independent Variable

a0= intercept of the line (Gives an additional degree of freedom)

a1 = Linear regression coefficient (scale factor to each input value).

ε = random error

The values for x and y variables are training datasets for Linear Regression model
representation.

12
Types of Linear Regression:

 Simple Linear Regression

 Multiple Linear Regression

 Logistic Regression:

A component of the supervised learning method is logistic regression. Using a specific

collection of independent variables, it aids us in predicting the results of categorical
dependent variables. Although it can be either binary (0 or 1) or boolean (true/false), it only
provides a probabilistic value between 0 and 1, not a precise value. Depending on how it is
used in the machine learning model, it is very comparable to linear regression. Similar to
how linear regression is used to address regression issues, logistic regression is useful in
addressing categorization issues.

The sigmoid function, which has the form of an S, can be used to represent logistic
regression. It foresees two highest numbers. (0 or 1).

Binomial, multinomial and Ordinal logistic regression models are examples.

 K Nearest Neighbour (KNN):

It belongs to the category of supervised learning methods and is one of the simplest machine
learning algorithms. Both classification and regression issues can be solved using it. It
thinks that the new data and the existing data are similar, and it places the new data in the
category that matches the existing categories the closest. Because it doesn't learn from the
training set right away but instead stores the dataset and executes an action on it when
classifying, this algorithm is also known as a lazy learner algorithm. Let's say we want to
determine whether a new picture is of a cat or dog. We have a few sets of images of cats
and dogs. Because KNN algorithm uses similarity measures, it is the best method for
identifying the cat from the data sets that are accessible. The KNN model will therefore
compare the new picture with the existing images and classify the output as belonging to
the cat.

13
 Fig: K Nearest Neighbour (KNN)

 K-Means Algorithm:

K-Means Clustering is a subpart of unsupervised learning techniques. By organizing the

unlabeled datasets into various groups, it aids in the solution of clustering problem. Here,
K specifies how many pre-defined clusters must be created as part of the procedure; for
example, if K=2, there will be two clusters, if K=3, there will be three clusters, and so on.

 Decision Tree:

Another type of machine learning method that falls under supervised learning is the decision
tree. The decision tree works similarly to KNN in that it aids in both classification and
regression problem solving, but it is typically used to address classification-related issues.
The classification algorithm is called a decision tree because it uses a classifier with a tree-
like structure, where characteristics are represented by internal nodes, decision rules are
represented by branches, and each leaf of the tree represents the classification result. The
decision node, also referred to as the root node, is where the tree begins, and the leaf node
is where it concludes.
While leaves are used to determine the results of those choices, decision nodes assist us in
14
making all decisions.
A decision tree is a visual representation of all possible solutions to an issue or choice based
on predetermined circumstances.

Fig: Decision Tree

 Random Forest:

One of the most popular machine learning methods that falls under the Supervised Learning
methodology is Random Forest. Similar to KNN and Decision Tree, it enables us to handle
classification and regression issues, but it is preferred whenever we need to address a
complex issue and enhance the performance of the model.
On the idea of ensemble learning, which is the process of combining various classifiers, a
random forest algorithm is built.
The random forest classifier is constructed using a variety of decision trees and different
dataset groups. This combination increases the model's accuracy by using input as the mean
forecast from all trees. Higher accuracy and overfitting are prevented by the larger number
of trees in the woodland. Additionally, it requires less training time than other methods.

 Support Vector Machine:

It is also among the most widely used machine learning methods, which are a subset of the
Supervised Learning approach. To readily place new data points in the appropriate category
in the future, the support vector machine algorithm aims to establish the best line or decision
boundary that can divide n-dimensional space into classes. A hyperplane is the name given
to this optimal judgment boundary. Both categorization and regression issues can be
resolved using it. It is used for tasks like text classification, image classification, and face

15
recognition.

 Naïve Bayes algorithm:

One of the most straightforward and efficient machine learning algorithms that falls under
the guided learning methodology is the naive Bayes algorithm. It is founded on the Bayes
Theorem, a theory that helps researchers solve problems involving classification. Building
quick machine learning models that can perform better and make quick forecasts is helpful.
Text classification with high-dimensional training datasets, it is primarily favored.

As a probabilistic classifier, it makes predictions based on the likelihood of an item. Some

significant uses of the Naive Bayes algorithm include spam filtering, sentiment analysis,
and article classification.

1.3 DEEP LEARNING

Deep learning is a type of machine learning and artificial intelligence that uses artificial
neural networks as its foundation. The term "deep" in deep learning alludes to the quantity
of layers in a neural network.

A group of algorithms known as "deep learning" were modeled after the structure and
operation of the human brain. It teaches computers effectively and predicts outcomes using
a vast quantity of structured and unstructured data. The way that data is presented is where
machine learning and deep learning methods diverge most. While deep learning uses neural
networks to build learning models, machine learning uses organized and unstructured data
to learn.

In machine learning, we must actively correct a model if it predicts incorrect outcomes.

Furthermore, these issues are automatically fixed by deep learning methods, so no explicit
action is required. One of the finest examples of deep learning is a self-driving car.

Natural language processing (NLP), image recognition, voice recognition, product

recommendation systems, and other complex issues can all be solved with the help of deep
learning.

The basic structure of deep learning:

16
 Fig: The basic structure of deep learning

Input layers, hidden layers, and output layers are just a few of the distinct layers found in
neural networks used in deep learning. The input layer takes in raw data, hidden layers are
used to extract any patterns and features that might be hidden in the data, and output
layers display the anticipated outcomes.

Deep learning is composed of a few easy stages.

1. Determine the weighted total.

2. Input the activation function in step 1 with this weighted total.

3. The activation function determines whether or not to activate the neuron and adds bias.

4. Make output predictions at the output layer.

5. Compare predicted and actual output, then use the backpropagation technique to
enhance the model's performance. The cost function is crucial in this stage for minimizing
the error rate.

 Types of Deep Neural Networks:

Deep learning networks come in a variety of forms, including Radial basis function neural
networks, feedforward neural networks, multi-layer perceptrons, convolution neural
networks (CNN), recurrent neural networks, modular neural networks, and sequence-to-
sequence models.

 Applications of Deep Learning:

17
Text, speech, and image recognition, self-driving cars, fraud detection, natural language
processing, virtual personal assistants, healthcare, infrastructure, banking & finance,
marketing, entertainment, and education are just a few of the sectors where deep learning
can be used. Playing games automatically, creating handwriting automatically, translating
languages automatically, restoring pixels in photos, labeling photos, predicting
demographics and elections, etc.

1.4 SIMILARITIES AND DIFFERENCES BETWEEN AI, ML & DL

Fig: Relationship between AI, ML & DL

For IT professionals, the words Deep Learning, Machine Learning, and Artificial
Intelligence are most frequently used on internet. All three of these tools are, however,
interconnected. Deep learning and machine learning can both be thought of as subsets of
artificial intelligence (AI). Alternatively, we could state that machine learning and deep
learning are both subsets of artificial intelligence.
Because of how similar these technologies appear, the majority of people believe that deep
learning, machine learning, and artificial intelligence are all interchangeable. But in
actuality, despite the fact that all of these technologies are used to create intelligent software
or machines that behave like humans, they still differ in terms of their features and
application.
It implies these three terms are often used interchangeably, but they do not quite refer to the
same things.

A subfield of computer science called artificial intelligence aids in the development of

intelligent, intelligent computers. Additionally, ML is a branch of AI that aids in machine
18
learning and the development of apps that use AI. On the other hand, deep learning is a
branch of machine learning that primarily uses neural networks while helping to build ML
models using massive amounts of data and complicated algorithms.

 Difference between AI, ML and DL:

Fig: Difference between AI, ML and DL

Artificial intelligence appeared around the 1950s. Machine learning emerged around the
1960s. Deep learning appeared around his 1970s.
AI stands for Machine Simulated Intelligence. Machine learning is a way to get machines
to make decisions without programming. Deep learning is the process of using artificial
neural networks to solve complex problems.
AI is a subset of data science. Machine learning is a subset of AI and data science. Deep
learning is a subset of machine learning, AI, and data science.
The purpose of AI is to build machines that can think like humans. The goal of machine
learning is to make machines learn from data so that they can solve problems. Deep His
learning goal is to build a neural network that automatically discovers patterns in feature
recognition.

1.5 SUMMARY

Basics of AI: Artificial Intelligence (AI) refers to the development of computer systems that can perform
tasks that typically require human intelligence. These tasks include reasoning, problem-solving, decision-
making, language understanding, and perception. AI can be divided into narrow AI, which is designed for
specific tasks, and general AI, which would possess human-like intelligence across various domains.

Basics of Machine Learning (ML): Machine Learning (ML) involves the use of algorithms that enable
computers to learn from data and improve their performance on a specific task over time. It's categorized
into supervised learning, where models learn from labeled data, and unsupervised learning, where models
19
identify patterns in unlabeled data. ML techniques find application in prediction, classification,
clustering, and recommendation systems.

Basics of Deep Learning: Deep Learning is a subset of machine learning that focuses on neural networks
with multiple layers. These deep neural networks can automatically learn hierarchical features from data,
allowing them to excel at tasks like image and speech recognition. Deep Learning has driven
breakthroughs in various AI domains, aided by its ability to automatically learn intricate patterns.

Similarities between AI, ML, and Deep Learning:

 All three fields are interconnected and contribute to the development of intelligent systems.

 Machine Learning and Deep Learning are subsets of Artificial Intelligence.

 They involve the use of data to make informed decisions or predictions.

Differences between AI, ML, and Deep Learning:

 AI encompasses a broader concept of machines emulating human intelligence, while ML and

Deep Learning are specific techniques within AI.

 ML focuses on algorithms that learn from data, while Deep Learning involves neural networks
with multiple layers to automatically learn complex features.

 Deep Learning is a subset of ML that specializes in neural networks with many layers.

1.6 KEYWORDS

AI: Artificial Intelligence

The broader concept of creating machines that can perform tasks that typically require human
intelligence, such as reasoning, problem-solving, and learning.

ML: Machine Learning

A subset of AI that involves developing algorithms and models that allow computers to learn from and
make predictions or decisions based on data.

DL: Deep Learning

A subset of machine learning that focuses on neural networks with many layers (deep neural networks) to
process and learn from data. It has shown remarkable success in various tasks, such as image and speech
recognition.

NN: Neural Network

A computational model inspired by the structure of the human brain, consisting of interconnected nodes
(neurons) organized in layers. Neural networks are the foundation of deep learning.

20
ANN: Artificial Neural Network

A type of neural network designed to mimic the behavior of biological neurons. It's the fundamental
building block of deep learning.

CNN: Convolutional Neural Network

A specialized type of neural network designed for processing grid-like data, such as images and videos.
CNNs use convolutional layers to automatically learn features from the input data.

RNN: Recurrent Neural Network

A type of neural network designed to handle sequences of data, such as time series or natural language.
RNNs have loops that allow information to be passed from one step of the sequence to the next.

MLP: Multi-Layer Perceptron

A basic type of feedforward neural network with multiple layers, including input, hidden, and output
layers. It's used for various tasks, such as classification and regression.

DL vs. ML:

Deep learning is a subset of machine learning that focuses on neural networks with many layers, enabling
it to automatically learn hierarchical features from data. ML encompasses a broader range of techniques
beyond deep learning, including decision trees, support vector machines, and more.

Supervised Learning:

A type of machine learning where the model is trained on labeled data, meaning that the input data is
paired with the correct output. The model learns to map inputs to outputs.

Unsupervised Learning:

A type of machine learning where the model is trained on unlabeled data, and the goal is to find patterns,
structures, or relationships within the data.

Reinforcement Learning:

A type of machine learning where an agent learns to interact with an environment to maximize a reward
signal. It learns through trial and error, making sequential decisions.

AI vs. ML vs. DL:

AI is the overarching concept of creating intelligent machines.

ML is a subset of AI that involves learning from data to make predictions or decisions.

DL is a subset of ML that focuses on deep neural networks for automatically learning features from data.

21
1.7 SELF-ASSESSMENT QUESTIONS

Long questions
1. What is the Difference Between AI, ML, and Deep Learning?
2. How Does Deep Learning Work and How Does It Relate to Neural Networks?

3. Explain the Supervised Learning Process with an Example.

4. Write a note on Real-World Applications of AI and Their Technological
Foundations.
5. Compare AI, ML & DL: Similarities and Differences.
6. Write a note on Unsupervised learning.

Short questions
1. What does AI stand for?
2. Define Machine Learning (ML).
3. What is Deep Learning?
4. Name one real-world application of AI.
5. Differentiate between supervised and unsupervised learning.
6. What is a neural network?

Fill in the blanks

1. Artificial Intelligence is about_____.
a) Playing a game on Computer
b) Making a machine Intelligent
c)Programming on Machine with your Own Intelligence
d)Putting your intelligence in Machine
2. A technique that was developed to determine whether a machine could or could not
demonstrate the artificial intelligence known as the___
a) Boolean Algebra b) Turing Test c) Logarithm d)Algorithm
3. Types of Artificial Intelligence is classified into ____ categories.
a) 4 b) 3 c) 2 d) 5
4. …………….. are still not developed, but researchers are making lots of efforts and
improvements for developing such AI machines.
a) Reactive Machines b) Limited Memory c)Theory of mind d)All of the above
5. Apple Siri is a good example of …………………….
a)Strong AI b) General AI c) Limited Memory d)Narrow AI
6. Machine learning is a subset of ___________.
a) Blockchain b) Artificial Intelligence c) Both A and B d) None of the above
22
7. Logistic regression is a ........... regression technique that is used to model data having a
............ outcome.
a)Linear, binary b)Linear, numeric c)Nonlinear, binary d)Nonlinear, numeric

8. A _________ is a decision support tool that uses a tree-like graph or model of decisions
and their possible consequences, including chance event outcomes, resource costs, and
utility.
a) Decision tree b) Graphs c) Trees d) Neural Networks
9.Random Forest has _________ as base learning models
a) multiple decision trees b) bagging c) Entropy d) none of these

True/ False
1.Regression discovers causal relationships.
a)True b)False

2. In the K-Means algorithm, we have to specify the number of clusters.

a)True b) False

Multiple choice question

1. What is the goal of Artificial Intelligence?
a) To solve artificial problems b) To extract scientific causes
c) To explain various sorts of intelligence d) To solve real-world problems
2. Which of the following is the branch of Artificial Intelligence?
a) Machine Learning b) Cyber forensics
c) Full-Stack Developer d) Network Design
3. The application/applications of Artificial Intelligence is/are:
Expert Systems b) Gaming c)Vision Systems d)All of the above
4. Which of the following is not an application of artificial intelligence?
Computer Vision b)Natural Language Processing
c)Database Management System d) Digital Assistants
5. Which of the following is an application of Artificial Intelligence?
a) It helps to exploit vulnerabilities to secure the firm
b) Language understanding and problem-solving (Text analytics and NLP)
c) Easy to create a website
d) It helps to deploy applications on the cloud
6. Who is known as the -Father of AI"?
a) Fisher Ada b) Alan Turing c)John McCarthy d)Allen Newell
7. Which of the given language is not commonly used for AI?
a) LISP b) PROLOG c)Python d)Perl

8. The component of an Expert system is_________.

23
a) Knowledge Base b) Inference Engine c) User Interface d)All of the above

9. Among the following option identify the one which is not a type of learning
a) Semi unsupervised learning b) Supervised learning
c) Reinforcement learning d) Unsupervised learning
10.What is machine learning ?
a) Machine learning is the science of getting computers to act without being explicitly
programmed.
b) Machine Learning is a Form of AI that Enables a System to Learn from Data.
c) Both A and B
d)None of the above
11. Application of Machine learning is __________.
a) email filtering b) sentimental analysis c) face recognition d) All
of the above

12. Which is the number of nearby neighbours to be used to classify the new record ?
a) KNN b)Validation data c)Euclidean Distance d)All the above

13. Support Vector Machine (SVM) can be used for _____.

a) classification only b) regression only c) classification and regression both d) None
of these
14. Naive Baye is?
a) Conditional Independence b) Conditional Dependence c) Both a and b d) None of
the above

15. What is representation in deep learning?

a) It is a way to look at data to represent or encode b) It gets closer to the expected
output
c) RGB and HSV are two different examples of representations d) All of the above
16.What is deep in deep learning?
a) The deep in deep learning is a reference to any kind of deeper understanding achieved
by the approach b) It stands for the idea of successive layers of representations in
deep learning
c) Answers A & B d) None of the above
17. Which statement is true?
a) Deep learning is a mathematical framework for learning representations from data
b) Deep learning is a biological framework for learning representations from brain
data
c) Deep learning is an analogue framework for learning representations from data
d) Deep learning is a digital framework for learning representations from data

18. What is relationship between Artificial intelligence (AI), machine learning (ML) and
deep learning(DL)?
a) There is no relationship between AI, ML, and DL
b) AI > ML > DL
24
c) DL > AI > ML
d) ML > AI > DL
19. Machine Learning is a field of AI consisting of learning algorithms that ..............
a) At executing some task b) Over time with experience c) Improve their
performance
d) All of the above

Match the following

Arrange the following stages of the machine learning process in the correct order:

1 Model Evaluation

2 Data Collection

3 Feature Selection

4 Model Training

5 Prediction

Answer keys:
Fill in the blanks
1.Artificial Intelligence is about_____.
b) Making a machine Intelligent

2. A technique that was developed to determine whether a machine could or could not
demonstrate the artificial intelligence known as the___
Ans: Turing Test

3. Types of Artificial Intelligence is classified into ____ categories.

Ans: 2
4. …………….. are still not developed, but researchers are making lots of efforts and
improvements for developing such AI machines.
Ans: Theory of mind
5. Apple Siri is a good example of …………………….
Ans: Narrow AI
6. Machine learning is a subset of ___________.
b) Artificial Intelligence
7. Logistic regression is a ........... regression technique that is used to model data having a
............ outcome.
c)Nonlinear, binary
25
8. A _________ is a decision support tool that uses a tree-like graph or model of decisions
and their possible consequences, including chance event outcomes, resource costs, and
utility.
a) Decision tree
9.Random Forest has _________ as base learning models
a) multiple decision trees

True / False
1.Regression discovers causal relationships.
a)True
2. In the K-Means algorithm, we have to specify the number of clusters.
a)True

Multiple Choice :
1. What is the goal of Artificial Intelligence?
c) To explain various sorts of intelligence
2. Which of the following is the branch of Artificial Intelligence?
a) Machine Learning
3. The application/applications of Artificial Intelligence is/are:
Ans. All of the above
4. Which of the following is not an application of artificial intelligence?
Ans. Database Management System
5. Which of the following is an application of Artificial Intelligence?
Ans. Language understanding and problem-solving (Text analytics and NLP)
6. Who is known as the -Father of AI"?
Ans: John McCarthy
7. Which of the given language is not commonly used for AI?
Ans: Perl
8. The component of an Expert system is_________.
Ans: All of the above
9. Among the following option identify the one which is not a type of learning
a) Semi unsupervised learning
10.What is machine learning ?
c) Both A and B
11. Application of Machine learning is __________.
d) All of the above
26
12. Which is the number of nearby neighbours to be used to classify the new record ?
a) KNN
13. Support Vector Machine (SVM) can be used for _____.
c) classification and regression both
14. Naive Baye is?
a) Conditional Independence
15. What is representation in deep learning?
d) All of the above
16.What is deep in deep learning?
b) It stands for the idea of successive layers of representations in deep learning
17. Which statement is true?
a) Deep learning is a mathematical framework for learning representations from data
18. What is relationship between Artificial intelligence (AI), machine learning (ML) and
deep learning(DL)?
a) AI > ML > DL
19. Machine Learning is a field of AI consisting of learning algorithms that ..............
d) All of the above

Match the following

1. Data Collection,
2. Feature Selection,
3. Model Training,
4. Model Evaluation,
5. Prediction

Long answers
1. What is the Difference Between AI, ML, and Deep Learning?
Answer - Refer section 1.4
2. How Does Deep Learning Work and How Does It Relate to Neural Networks?
Answer - Refer section 1.3
3. Explain the Supervised Learning Process with an Example.
Answer - Refer section 1.2.1
4. Write a note on Real-World Applications of AI and Their Technological
Foundations.
Answer - Refer section 1.1.1
5. Compare AI, ML & DL: Similarities and Differences.
Answer - Refer section 1.4
6. Write a note on Unsupervised learning.

27
 Answer - Refer section 1.2.1

Short questions
1. What does AI stand for?
Answer - Refer section 1.1
2. Define Machine Learning (ML).
Answer - Refer section 1.2
3. What is Deep Learning?
Answer - Refer section 1.3
4. Name one real-world application of AI.
Answer - Refer section 1.1.1
5. Differentiate between supervised and unsupervised learning.
Answer - Refer section 1.2.1
6. What is a neural network?
Answer - Refer section 1.3

Suggested Reading:
Books:
- "Artificial Intelligence: A Modern Approach" by Stuart Russell and Peter Norvig
This comprehensive book provides an introduction to various aspects of artificial
intelligence, including machine learning and deep learning, and covers their similarities and
differences.
- "Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow" by Aurélien
Géron
A practical guide that takes you through the fundamentals of machine learning, deep
learning, and neural networks using popular libraries like Scikit-Learn, Keras, and
TensorFlow.
-"Deep Learning" by Ian Goodfellow, Yoshua Bengio, and Aaron Courville
This book is a comprehensive resource for understanding deep learning concepts,
architectures, and algorithms. It covers the foundations of neural networks and their
applications.

28