Artificial Intelligence

(AI)
AGENDA

The Agenda of this presentation is:

▰ What is Machine Learning..?

▰ Supervised Machine Learning
▰ Unsupervised Machine Learning
▰ Semi Supervised Machine Learning
▰ Reinforcement Learning

2
What is Machine Learning ?

▰ Machine Learning is making the computer learn from studying data

and statistics.
▰ Machine Learning is a step into the direction of artificial intelligence
(AI).
▰ Machine Learning is a program that analyses data and learns to
predict the outcome
 Machine Learning Process

Step1 • Define the objective of the Problem

Statement

Step 2 • Data Gathering

Step 3 • Data Preparation

Step 4 • Exploratory Data Analysis

Step 5 • Building a Machine Learning Model

Step 6 • Model Evaluation & Optimization

Step 7 • Predictions
4
Machine Learning Types

5
Supervised Vs
Unsupervised Learning

6
 Learnings
Learning:
• Building black of AI.
• Method of improving
an AI program's
knowledge by making
observations about
its surroundings.
Semi- Re-
Supervised Unsupervised Supervised enforcement
Learning Learning Learning Learning

7
Back to main
 Supervised
Learning

8
Back to main
 Supervised Learning

❑ Machine learning approach that uses labelled datasets.

❑ Some data tagged with correct answer.
❑ Helps to predict/forecast outcomes for unexpected data.
❑ At least some of this approach involves human control. The vast bulk of data is unlabeled
and unprocessed.
❑ Models need to define mapping function to map input variable (X) with the output variable
(Y).
❑ Y=f(X)
❑ Organizations use to tackle a range of real-world problems, such as spam classification in a
distinct folder from your email.

9
Why Supervised Learning?

❖ Collect data or create a data output from past experience

with supervised learning.
❖ With the help of experience, it is possible to improve
performance criteria.
❖ Use supervised machine learning to tackle a variety of
real-world situations.

10
How Supervised Learning works?

❑ each models to produce the desired output.

❑ training dataset contains both correct and incorrect outputs, allowing the
model to improve over time.
❑ Loss function used to assess the algorithm's correctness and adjusted
until the error is suitably minimized.

Learning Phase:

Training Clean, prepare & Algorithm Model

Data manipulate data

11
Unsupervised Learning

❑ Machine learning approach that uses unlabeled datasets.

❑ Some data tagged with incorrect answer.
❑ Helps to predict/forecast outcomes for unexpected data.
❑ Algorithms discover hidden patterns in data without the need for human intervention.
❑ suited to answer questions about unseen trends and relationships within data itself.
❑ Allow you to perform more complex processing tasks.

https://datasciencetoday.net/

12
Why Unsupervised Learning?

❖ Finds all kind of unknown patterns in data.

❖ Find features which can be useful for categorization.
❖ Taken place in real time, so all input data to be analyzed and labeled in the
presence of learners.
❖ Easier to get unlabeled data from a computer than labeled data, which needs
manual intervention.

https://datasciencetoday.net/

13
How Unsupervised Learning works?

❑ Works by analyzing uncategorized, unlabeled data and finding hidden

structures in it.
❑ Learning process begins with the data scientists training the algorithms
using the training datasets.
❑ Data points in these datasets are unlabeled and uncategorized

Learning Phase:

Test Data Improve/ Validate Data

Clean, prepare & Train Model
Get Data
manipulate data

14
Categories

Two types of categories

Clustering Association

15
 Clustering

❖ Clustering:
➢ Data mining technique for grouping unlabeled data based on their similarities or differences.
➢ Mainly deals with finding a structure or pattern in a collection of uncategorized data.
➢ process data and find natural clusters(groups) if they exist in the data.
➢ Modify how many clusters pf algorithms should identify also allows to adjust the granularity of these
groups.
Examples:

https://www.istockphoto.com/
Market segmentation Identify Fake News 16
 Association

❖ Association:
➢ Allow to establish associations amongst data objects inside large databases.
➢ Discovering exciting relationships between variables in large databases.
Examples:
➢ people that buy a new home most likely to buy new furniture.
➢ In market basket analysis and recommendation engines, along the lines of “Customers Who Bought
food Item Pizza Also Bought” coke as a recommendations .

https://blog.vantagecircle.com/ https://www.redbubble.com/

Buy new house also buy new furniture Pizza coke recommendations 17
Examples

Google News
• Use different clusters for
same data about news
articles
Examples

Image credit: NASA/JPL-Caltech/E. Churchwell (Univ. of Wisconsin, Madison)

Social network analysis Astronomical data analysis

Class Task

following examples, which would you address using an

unsupervised learning algorithm? (Check all that apply.)

➢ Given email labeled as spam/not spam, learn a spam filter.

➢ Given a set of news articles found on the web, group them into
set of articles about the same story.
➢ Given a database of customer data, automatically discover market
segments and group customers into different market segments.
➢ Given a dataset of patients diagnosed as either having diabetes
or not, learn to classify new patients as having diabetes or not.
20
 Supervised VS Unsupervised Learning
Supervised machine learning Unsupervised machine
Parameters
technique learning technique
Input and output variables will be
Process only input data will be given
given.

Algorithms are trained using labeled Algorithms are used against data
Input Data
data. which is not labeled
Support vector machine, Neural
divided into different categories: like
network, Linear and logistics
Algorithms Used Cluster algorithms, K-means,
regression, random forest, and
Hierarchical clustering, etc.
Classification trees.

Computational Complexity simpler method. computationally complex

Highly accurate and trustworthy Less accurate and trustworthy

Accuracy of Results
method. method.
Number of Classes Number of classes is known. Number of classes is not known. Back to main

cannot get precise information

Classifying big data can be a real regarding data sorting, and the
Main Drawback
challenge in Supervised Learning. output as data used in unsupervised 21
learning is labeled and not known.
 Semi-supervised
Learning

22
Back to main
Semi-supervised Learning

❑ combination of both supervised and unsupervised learning.

❑ Involves small number of labeled examples and large number of unlabeled examples.
❑ Provides benefits of both supervised and unsupervised learning.
❑ An ideal for medical images, where small amount of training data can lead to
significant improvement in accuracy.
❑ Helps to predict/forecast outcomes more accurately.

For example,
Radiologist label small subset of CT scans for tumors or diseases so the machine can
more accurately predict which patients might require more medical attention.

https://datasciencetoday.net/

23
 Real-world applications of Semi-supervised
Learning

▰ Speech Analysis ▰ Protein

Sequence
Classification

https://www.toolbox.com/ https://www.ebi.ac.uk/

▰ Text Document
▰ Web Content
Classifier
Classification
Back to main

https://nanonets.com/
24
https://towardsdatascience.com/
Reinforcement Learning

❑ machine learning training method based on rewarding desired

behaviors and/or punishing undesired ones.
❑ In general, agent is able to perceive and interpret its
environment.
❑ Take actions and learn through trial and error.
❑ Assigns positive values to desired actions to encourage the
agent and negative values to undesired behaviors.
❑ This method has been adopted in artificial intelligence (AI) as a
way of directing unsupervised machine learning through
rewards and penalties.
❑ There is no answer but the reinforcement agent decides what
https://en.wikipedia.org/
to do to perform the given task.
❑ In the absence of a training dataset, it is bound to learn from its
experience.
25
Back to main
Example:
The problem is as follows: We have an agent and a reward, with many
hurdles in between. The agent is supposed to find the best possible path to
reach the reward. The following problem explains the problem more easily

Back to main

26
https://www.geeksforgeeks.org/