Ai Module 3

Amity School of Engineering & Technology
Classical programming uses programs (algorithms) to create

results:
Traditional Computing
Data + Computer Algorithm = Result
Machine Learning uses results to create programs (algorithms):
Machine Learning
Data + Result = Computer Algorithm
1
Machine Learning
Machine Learning is often considered equivalent with Artificial
Intelligence.
This is not correct. Machine learning is a subset of Artificial Intelligence.
Machine Learning is a discipline of AI that uses data to teach machines.
Machine Learning is a field of study that gives computers

the ability to learn without being programmed."
Machine learning is the subfield of AI that focuses on the development of the computer
programs which have access to data by providing a system with the ability to learn and
improve automatically.
2
Intelligent Decision Formula

• Save the result of all actions
• Simulate all possible outcomes
• Compare the new action with the old ones
• Check if the new action is good or bad
• Choose the new action if it is less bad
• Do it all over again
The fact that computers can do this millions of times, has proven that
computers can take very intelligent decisions.
3
Supervised Learning
Supervised learning uses labeled data (data with known answers) to
train algorithms to:
• Classify Data
• Predict Outcomes
Supervised learning can classify data like "What is spam in an e-mail",

based on known spam examples.
Supervised learning can predict outcomes like predicting what kind of
video you like, based on the videos you have played.
Example :
• Market Prediction – It is also called Regression. Historical business market data is
fed to the computer. Then, with analysis and regression algorithm, the new price for
the future is predicted depending on variables.
• Image Classification – The algorithm is drawn from feeding with labeled image data.
An algorithm is trained, and it is expected that the algorithm classifies it correctly in
the case of the new image.
4
Unsupervised Learning
Unsupervised learning is used to predict undefined relationships like
meaningful patterns in data.
It is about creating computer algorithms than can improve themselves.
It is expected that machine learning will shift to unsupervised learning
to allow programmers to solve problems without creating models.
Example:
• Clustering – Data with similar traits are asked to group together by the algorithm; this
grouping is called clusters. These prove helpful in the study of these groups, which can
be applied to the entire data within a cluster more or less.
• High Dimension Data – High dimension data is normally not easy to work with. With
the help of unsupervised learning, visualization of high dimension data becomes
possible.
• Generative Models – Once your algorithm analyses and comes up with the probability
distribution of the input, it can be used to generate new data. This proves to be very
helpful in cases of missing data.
5
Self-Supervised Learning
Self-supervised learning is similar to unsupervised learning because it
works with data without human added labels.
The difference is that unsupervised learning uses clustering, grouping,
and dimensionality reduction, while self-supervised learning draw its own
conclusions for regression and classification tasks.
Reinforcement Machine Learning

This type of machine learning algorithm uses the trial and error method to churn out output
based on the highest efficiency of the function. The output is compared to find out errors and
feedback fed back to the system to improve or maximize its performance. The model is
provided with rewards which are basically feedback and punishments in its operations while
performing a particular goal.
6
7
Semi-Supervised Machine Learning

These algorithms normally undertake labeled and unlabeled data, where the unlabelled data
amount is large as compared to labeled data. As it works with both and in between
supervised and unsupervised learning algorithms, therefore is called semi-supervised
machine learning. Systems using these models are seen to have improved learning accuracy.
Example – An image archive can contain only some of its data labeled, e.g. Dog, cat,
mouse, and a large chunk of images remain unlabelled.
8
Below are the types of Machine learning models based on the kind of outputs we expect from
the algorithms:
9
1. Classification
There is a division of classes of the inputs; the system produces a model from training data
wherein it assigns new inputs to one of these classes. It falls under the umbrella of
supervised learning. A real-life example can be spam filtering, where emails are the input
that is classified as “spam” or “not spammed”.
2. Regression
Regression algorithm also is a part of supervised learning, but the difference being that the
outputs are continuous variables and not discrete.
Example – Predicting house prices using past data.
3. Dimensionality Reduction
This type of Machine Learning is related to analyses of inputs and reduces them to only
relevant ones for model development. Feature selection, i.e. input selection and feature
extraction, is further topics needed to better understand dimensionality reduction.
On the basis of the above different approaches, there are various algorithms to be
considered. Some very common algorithms being Linear and Logistic Regression, K-
nearest neighbors, Decision trees, Support vector machines, Random Forest, etc. With the
help of these algorithms, complex decision problems can have a sense of direction based
on a huge amount of data. In order to attain this accuracy and opportunities, added
resources, as well as time, are required to be provided. Therefore, machine learning used
along with Artificial intelligence and other technologies is more effective to process
information.
10
Most Common Types of Machine Learning Problems
these problem types and understand the details.
• Classification
• Clustering
• Time-series forecasting
• Anomaly detection
• Ranking
• Recommendation
• Data generation
• Optimization
• Regression
11
Regression
When the need is to predict numerical values, such kinds of problems are
called regression problems. For example, house price prediction
Linear regression, K-NN, random forest, neural networks
Classification
When there is a need to classify the data in different classes, it is called a

classification problem. If there are two classes, it is called a binary
classification problem. When it is multiple classes, it is multi-nomial
classification. For example, classify whether a person is suffering from a
disease or otherwise. Classify whether a stock is “buy”, “sell”, or “hold”.
Check this related post – Machine learning techniques for stock prediction
Logistic regression, random forest, K-NN, gradient boosting

classifier, neural networks
12
Clustering
When there is a need to categorize the data points in similar groupings or

clusters, this is called a clustering problem.
-Means, DBSCAN, Hierarchical clustering, Gaussian mixture

models, BIRCH
13
Time-series forecasting
When there is a need to predict a number based on the time-series data, it is

called a time-series forecasting problem. A time series is a sequence of
numerical data points in successive order. Time series data means that data is
in a series of particular time periods or intervals. For example, a time-series
forecasting problem is about forecasting the sales demand for a product,
based on a set of input data such as previous sales figures, consumer
sentiment, and weather. Another kind of time series problem is demand
forecasting.
RIMA, SARIMA, LSTM, Exponential smoothing, Prophet,

GARCH, TBATS, Dynamic linear models
14
Anomaly detection
When there is a need to find the outliers in the dataset, the problem is called an
anomaly detection problem. In other words, if a given record can be classified as
an outlier or unexpected event/item, this can be called an anomaly detection
problem. For example, credit card fraud transactions detection is an anomaly
detection problem.
IsolationForest, Minimum covariance determinant, Local outlier

factor, One-class SVM
15
Ranking
When there is a need to order the results of a request or a query based on some
criteria, the problem is ranking problems. We rank the output of query
execution based on scores we assign to each output based on some algorithms.
These algorithms are called a ranking algorithm. Recommendation engines
make use of the ranking algorithm to recommend the next items.
BIpartite ranking (Bipartite Rankboost,

Bipartite RankSVM)
16
Recommendation
When there is a need to recommend such as “next item” to buy or “next video”
to watch or “next song” to listen to, the problem is called a recommendation
problem. The solutions to such problems are called recommender systems.
Content-based and collaborative filtering machine

learning methods
17
Data generation
When there is a need to generate data such as images, videos, articles,
posts, etc, the problem is called a data generation problem.
Generative adversarial network (GAN), Hidden Markov

models
Optimization
When there is a need to generate a set of outputs that optimize

outcomes related to some objective (objective function), the problem is
called an objective function.
Linear programming methods, genetic

programming
18
Machine learning is a subset of Artificial Intelligence (AI) which uses algorithms to

learn from data. If you are familiar with AI as it has been through the news
recently in regards to fending off cyber attacks or causing self-driving cars
accidents then this will be easy for you understand. The idea that machines can
“learn” without being programmed by humans goes back further than just recent
years though; think about the first computers in the history of machine learning
and how they could only do one thing at a time.
One of the main reasons that machine learning is important for business use
today is because it helps businesses be more efficient in their processes, as well as
provide better customer service through AI-assisted chat bots or automated
emails. It also provides great tools to help teach other people about new things
like geography or historical events!
19
20
1943: The First Neutral Network with Electric Circuit

The first neutral network with electric circuit was developed by Warren McCulloch
and Walter Pitts in 1943. The goal of the network was to solve a problem that had
been posed by John von Neumann and others: how could computers be made to
communicate with each other?
This early model showed that it was possible for two computers to communicate
without any human interaction. This event is important because it paved the way
for machine learning development.
1950: Turing Test
The Turing Test is a test of artificial intelligence proposed by mathematician Alan
Turing. It involves determining whether a machine can act like a human, or if
humans can’t tell the difference between human and machine given answers.
The goal of the test is to determine whether machines can think intelligently and
demonstrate some form of emotional capability. It does not matter whether the
answer is true or false but whether it is considered human or not by the
questioner. There have been several attempts to create an AI that passes the
Turing Test, but no machine has yet successfully done so.
The Turing Test has been criticized because it measures how much a machine can
imitate a human rather than proving their true intelligence.
21
• 1952: Computer Checkers

• Arthur Samuel was a pioneer in machine learning and is credited with
creating the first computer program to play championship-level checkers.
His program, which he developed in 1952, used a technique called alpha-
beta pruning to measure the chances of winning a game. This method is
still widely used in games today. In addition, Samuel also developed the
minimax algorithm, which is a technique for minimizing losses in games.
• 1957: Frank Rosenblatt – The Perceptron
• Frank Rosenblatt was a psychologist who is most famous for his work on
machine learning. In 1957, he developed the perceptron, which is a
machine learning algorithm. The Perceptron was one of the first
algorithms to use artificial neural networks, widely used in machine
learning.
• It was designed to improve the accuracy of computer predictions. The
goal of the Perceptron was to learn from data by adjusting its parameters
until it reached an optimal solution. Perceptron’s purpose was to make it
easier for computers to learn from data and to improve upon previous
methods that had limited success.
22
• 1967: The Nearest Neighbor Algorithm

• The Nearest Neighbor Algorithm was developed as a way to
automatically identify patterns within large datasets. The goal of this
algorithm is to find similarities between two items and determine which
one is closer to the pattern found in the other item. This can be used for
things like finding relationships between different pieces of data or
predicting future events based on past events.
• In 1967, Cover and Hart published an article on “Nearest neighbor
pattern classification.” It is a method of inductive logic used in machine
learning to classify an input object into one of two categories. The
pattern classifies the same items that are classified in the same
categories as its nearest neighbors. This method is used to classify
objects with a number of attributes, many of which are categorical or
numerical and may have overlapping values.
23
• 1974: The Backpropagation

• Backpropagation was initially designed to help neural networks
learn how to recognize patterns. However, it has also been used in
other areas of machine learning, such as boosting performance
and generalizing from data sets to new instances. The goal of
backpropagation is to improve the accuracy of a model by
adjusting its weights so that it can more accurately predict future
outputs.
• Paul Werbos laid the foundation for this approach to machine
learning in his dissertation in 1974, which is included in the book “
The Roots of Backpropagation“.
• 1979: The Stanford Cart
• The Stanford Cart is a remote-controlled robot that can move
independently in space. It was first developed in the 1960s and
reached an important milestone in its development in 1979. The
purpose of the Stanford Cart is to avoid obstacles and reach a
specific destination: In 1979, “The Cart” succeeded for the first time
in traversing a room filled with chairs in 5 hours without human
24
25
• Present: State-of-the-art Machine Learning

• Machine learning is used in many different fields,
from fashion to agriculture. Machine Learning
algorithms are able to learn patterns and
relationships between data, find predictive insights
for complex problems and extract information that is
otherwise too difficult to find. Today’s Machine
Learning algorithms are able to handle large amounts
of data with accuracy in a relatively short amount of
time.
26
• 1997: A Machine Defeats a Man in Chess

• In 1997, the IBM supercomputer Deep Blue defeated chess grandmaster
Garry Kasparov in a match. It was the first time a machine had beaten
an expert player at chess and it caused great concern for humans in the
chess community. This was a landmark event as it showed that AI
systems could surpass human understanding in complex tasks.
• This marked a magical turning point in machine learning because the
world now knew that mankind had created its own opponent- an
artificial intelligence that could learn and evolve on its own.
• 2002: Software Library Torch
• Torch is a software library for machine learning and data science. Torch
was created by Geoffrey Hinton, Pedro Domingos, and Andrew Ng to
develop the first large-scale free machine learning platform. In 2002, the
founders of Torch created it as an alternative to other libraries because
they believed that their specific needs were not met by other libraries.
As of 2018, it has over 1 million downloads on Github and is one of the
most popular machine learning libraries available today.
• Keep in mind: No longer in active development, however, PyTorch can
be used, which is based on the Torch Library.
27
• 2006: Geoffrey Hinton, the father of Deep Learning
• In 2006, Geoffrey Hinton published his “A Fast Learning Algorithm for Deep Belief
Nets.” This paper was the birth of deep learning. He showed that by using a deep
belief network, a computer could be trained to recognize patterns in images.
• Hinton’s paper described the first deep learning algorithm that can achieve
human-level performance on difficult and complex pattern recognition tasks.
• 2011: Google Brain
• Google Brain is a research group of Google devoted to artificial intelligence and
machine learning. The group was founded in 2011 by Google X and is located in
Mountain View, California. The team works closely with other AI research groups
within Google such as the DeepMind group that has developed AlphaGo, an AI
that defeated the world champion at Go. Their goal is to build machines that can
learn from data, understand language, answer questions in natural language, and
have common sense reasoning.
• The group is, as of 2021, led by Geoffrey Hinton, Jeff Dean and Zoubin
Ghahramani and focuses on deep learning, a model of artificial neural networks
that is capable to learn complex patterns from data automatically without being
explicitly programmed.
28
• 2014: DeepFace
• DeepFace is a deep learning algorithm which was originally
developed in 2014 and is part of the company “Meta”. The
project received significant media attention after it
outperformed human performance on the well-known “Faces in
the Wild” test.
• DeepFace is based on a deep neural network, which consists of
many layers of artificial neurons and weights that connect each
layer to its neighboring ones. The algorithm takes as input a
training data set of photographs, with each photo annotated
with the identity and age of its subject. The team has been very
successful in recent years and published many papers on their
research results. They have also trained several deep neural
networks that have achieved significant success in pattern
recognition and machine learning tasks.
29
• 2017: ImageNet Challenge – Milestone in the History of

Machine Learning
• The ImageNet Challenge is a competition in computer vision
that has been running since 2010. The challenge focuses on the
abilities of programs to process patterns in images and
recognize objects with varying degrees of detail.
• In 2017, a milestone was reached.
29 out of 38 teams achieved 95% accuracy with their computer
vision models. The improvement in image recognition is
immense.
30
31
Design a Learning System in
Machine Learning
• According to Arthur Samuel “Machine Learning enables a
Machine to Automatically learn from Data, Improve
performance from an Experience and predict things without
explicitly programmed.”
• In Simple Words, When we fed the Training Data to Machine
Learning Algorithm, this algorithm will produce a
mathematical model and with the help of the mathematical
model, the machine will make a prediction and take a decision
without being explicitly programmed. Also, during training
data, the more machine will work with it the more it will get
experience and the more efficient result is produced.
32
33
• Example : In Driverless Car, the training data

is fed to Algorithm like how to Drive Car in
Highway, Busy and Narrow Street with factors
like speed limit, parking, stop at signal etc.
After that, a Logical and Mathematical model
is created on the basis of that and after that, the
car will work according to the logical model.
Also, the more data the data is fed the more
efficient output is produced.
34
• Designing a Learning System in Machine Learning :

• According to Tom Mitchell, “A computer program is said to be
learning from experience (E), with respect to some task (T).
Thus, the performance measure (P) is the performance at task
T, which is measured by P, and it improves with experience
E.”
• Example: In Spam E-Mail detection,
• Task, T: To classify mails into Spam or Not Spam.
• Performance measure, P: Total percent of mails being
correctly classified as being “Spam” or “Not Spam”.
• Experience, E: Set of Mails with label “Spam”
35
Steps for Designing Learning
System are:
36
• Step 1) Choosing the Training

Experience: The very important and first task
is to choose the training data or training
experience which will be fed to the Machine
Learning Algorithm. It is important to note that
the data or experience that we fed to the
algorithm must have a significant impact on
the Success or Failure of the Model. So
Training data or experience should be chosen
wisely.
37
• Step 2- Choosing target function: The next

important step is choosing the target function. It
means according to the knowledge fed to the
algorithm the machine learning will choose
NextMove function which will describe what type of
legal moves should be taken. For example : While
playing chess with the opponent, when opponent will
play then the machine learning algorithm will decide
what be the number of possible legal moves taken in
order to get success.
38
• Step 3- Choosing Representation for Target

function: When the machine algorithm will know all
the possible legal moves the next step is to choose the
optimized move using any representation i.e. using
linear Equations, Hierarchical Graph Representation,
Tabular form etc. The NextMove function will move
the Target move like out of these move which will
provide more success rate. For Example : while
playing chess machine have 4 possible moves, so the
machine will choose that optimized move which will
provide success to it.
39
• Step 4- Choosing Function Approximation Algorithm: An

optimized move cannot be chosen just with the training data.
The training data had to go through with set of example and
through these examples the training data will approximates
which steps are chosen and after that machine will provide
feedback on it. For Example : When a training data of Playing
chess is fed to algorithm so at that time it is not machine
algorithm will fail or get success and again from that failure or
success it will measure while next move what step should be
chosen and what is its success rate.
40
• Step5) Final Design: The final design is

created at last when system goes from number
of examples , failures and success , correct
and incorrect decision and what will be the
next step etc. Example: DeepBlue is an
intelligent computer which is ML-based won
chess game against the chess expert Garry
Kasparov, and it became the first computer
which had beaten a human chess expert.
41

Ai Module 3

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ai Module 3

Uploaded by

Copyright:

Available Formats

Amity School of Engineering & Technology

Classical programming uses programs (algorithms) to create

Machine Learning is a field of study that gives computers

Intelligent Decision Formula

Supervised learning can classify data like "What is spam in an e-mail",

Reinforcement Machine Learning

Semi-Supervised Machine Learning

Most Common Types of Machine Learning Problems

these problem types and understand the details.

Linear regression, K-NN, random forest, neural networks

When there is a need to classify the data in different classes, it is called a

Logistic regression, random forest, K-NN, gradient boosting

When there is a need to categorize the data points in similar groupings or

-Means, DBSCAN, Hierarchical clustering, Gaussian mixture

When there is a need to predict a number based on the time-series data, it is

RIMA, SARIMA, LSTM, Exponential smoothing, Prophet,

IsolationForest, Minimum covariance determinant, Local outlier

BIpartite ranking (Bipartite Rankboost,

Content-based and collaborative filtering machine

Generative adversarial network (GAN), Hidden Markov

When there is a need to generate a set of outputs that optimize

Linear programming methods, genetic

Machine learning is a subset of Artificial Intelligence (AI) which uses algorithms to

1943: The First Neutral Network with Electric Circuit

• 1952: Computer Checkers

• 1967: The Nearest Neighbor Algorithm

• 1974: The Backpropagation

• Present: State-of-the-art Machine Learning

• 1997: A Machine Defeats a Man in Chess

• 2017: ImageNet Challenge – Milestone in the History of

• Example : In Driverless Car, the training data

• Designing a Learning System in Machine Learning :

• Step 1) Choosing the Training

• Step 2- Choosing target function: The next

• Step 3- Choosing Representation for Target

• Step 4- Choosing Function Approximation Algorithm: An

• Step5) Final Design: The final design is

You might also like