Artificial

Intelligence
How do we define
Artificial
Intelligence?
 ARTIFICIAL + INTELLIGENCE
refers to the ability to
refers to something that is
learn, understand,
made or created by
reason, and solve
humans rather than
problems.
occurring naturally
 INFORMAL AND EPHEMERAL DEFINITION
“AI is the study of how to make
computers do things, which, at the
moment, people do better.”
 Formal and Official Definition

• Artificial Intelligence refers to the simulation of

human intelligence in machines that are
programmed to think, learn, and perform tasks
typically requiring human intelligence.

• AI enables machines to perform tasks such as

problem-solving, and decision-making.
History of AI
Milestones in AI Development

Rise of
The Turing The Deep
ELIZA IBM Watson Generative
Test Blue
AI
1950 1966 1997 2011 2020

Alan Turing An early defeats world wins a quiz AI

introduces the chatbot chess show while advancements
Turing Test to developed to champion outperforming in self-driving
determine if a mimic Garry human technology,
machine can conversation Kasparov. champions. healthcare,
exhibit human- by recognizing and NLP with
like keywords tools like GPT-
intelligence. 4o, Llama.
What is an AI Technique?
Defining the problem as a State
Space?
Solution to
Water Jug
Problem
 Problem
Characteristics
 What is an
Expert System?
 Tasks of an
Expert System
Characteristics
of an Expert
System
Architecture of
Expert Systems
Knowledge
Acquisition
Knowledge
Elicitation
 Stages of
Knowledge
Acquisition
 Representing
the Knowledge
Rules as a Knowledge
Representation
Technique
The Inference
Engine
Additional
Problems
How do we define
Data Mining?
Formal Definition
Data mining is the process of discovering
interesting patterns and knowledge from
large amounts of data. The data sources
can include databases, data warehouses, the
Web, other information repositories, or data
that are streamed into the system dynamically.
Process of Knowledge Discovery
from Data
1. Data Cleaning: To remove noise and inconsistent data.
2. Data Integration: Where multiple data sources may be combined.
3. Data selection: Where data relevant to the analysis task are retrieved
from the database.
4. Data transformation: Where data are transformed and consolidated
into forms appropriate for mining by performing summary or aggregation
operations.
5. Data Mining: an essential process where intelligent methods are applied
to extract data patterns).
6. Pattern Evaluation: to identify the truly interesting patterns
representing knowledge based on interestingness measures.
7. Knowledge presentation: where visualization and knowledge
representation techniques are used to present mined knowledge to users.
 General Process of
Knowledge Discovery from
Data
 Representation of
Classification using different
types of data modelling
Cluster Analysis of Random
Data
 Terms about Interestingness about
Data
1. Outlier Analysis
• A data set may contain objects that do not comply with the
general behavior or model of the data. These data objects are
outliers. Many data mining methods discard outliers as noise or
exceptions. In some applications e.g., fraud detection the rare
events can be more interesting than the more regularly
occurring ones. The analysis of outlier data is termed as outlier
analysis or anomaly mining.

2. Interestingness of a Pattern in Data Mining

• Not all patterns are interesting, only the ones that have a particular
threshold of support and confidence are interesting.
Data Mining is an Interdisciplinary
Field
 Disciplines of Data Mining
1. Statistics
a. Focuses on data collection, analysis, and presentation.
b. Helps build statistical models for data classification, pattern recognition, and
prediction.
2. Machine Learning
a. Enables computers to learn patterns and make decisions from data.
b. Includes supervised (classification), unsupervised (clustering), and semi-
supervised learning.
3. Database Systems and Data Warehouses
a. Manages large-scale structured data efficiently.
b. Offers scalability for large datasets, often through data warehouses.
c. Enables data mining by integrating data from multiple sources.
4. Information Retrieval
a. Searches for unstructured information, often using probabilistic models.
b. Useful for text and multimedia mining, especially on large data collections like the
Web.
 Applications Targeted by Data
Mining
1. Business Intelligence (BI):
a. BI helps businesses understand customer behavior, market trends,
competitors, and operations.
b. Core techniques include classification, prediction, clustering, and OLAP for
sales, customer management, and market insights.
2. Web Search Engines
a. Search engines use data mining to handle large-scale web data,
including crawling, indexing, and ranking search results.
b. Data mining improves query classification, personalized
recommendations, and context-aware search results.
 Issues in Data Mining
1. Mining Methodology - Develop new methods to handle
diverse knowledge types, data uncertainty, and
multidimensional data mining.
2. User Interaction - Enhance interactive mining, incorporate
user knowledge, and improve result visualization.
3. Efficiency and Scalability - Ensure mining algorithms are
scalable, efficient, and capable of handling large, distributed
data.
4. Diversity of Data Types - Address the challenges of mining
structured, semi-structured, and unstructured data across
diverse sources.
5. Data Mining and Society - Focus on social impact, privacy
concerns, and seamless integration of mining into daily life.
 Types of Data Attributes
An attribute is a data field representing a characteristic or feature of a
data object.
1. Nominal Attributes are categorical data with values that represent
names or categories like hair color or marital status. No meaningful
order, mathematical operations don't apply.
2. Special case of nominal data with only two possible values is
Binary Data, like true/false. These can be symmetric both states are
equally important (gender) or asymmetric one state is more
important (cancer).
3. Ordinal Values have ordered values such as sizes like small,
medium, large, but the difference between successive values is
unknown. They are useful for ranking but not for calculating averages.
4. Quantitative attributes with purely measurable values either
interval-scaled or ratio-scaled. Allow all mathematical operations.
Advantages of Data Mining
1. Data mining excels at identifying patterns and trends within
large datasets, which can lead to valuable insights for decision-
making processes across various sectors such as healthcare, finance,
and marketing.
2. Enables organizations to predict future trends based on historical
data. For instance, businesses can forecast customer purchasing
behaviors, allowing for targeted marketing strategies (Market Basket
Analysis).
3. By automating the data analysis process data mining enhances
efficiency.
4. Companies can gain a deeper understanding of customer preferences
and behaviors enabling personalized marketing efforts that improve
customer satisfaction and loyalty.
Disadvantages of Data Mining
1. The effectiveness of data mining is highly dependent on the
quality of the data used. Poor quality or biased data can lead
to inaccurate predictions and misleading insights.
2. The use of personal data raises ethical issues regarding
privacy. Organizations must navigate legal frameworks and
ethical considerations when mining sensitive information.
3. There is a risk of overfitting models to training data.
4. Implementing data mining techniques requires significant
expertise in statistics and machine learning.
Resources
• Kevin Knight, Elaine Rich, B. Nair - Artificial
Intelligence - Tata McGraw-Hill Education Pvt. Ltd.
(2010)
• Jiawei Han, Jian Pei, Hanghang Tong - Data
Mining Concepts and Techniques - The Morgan
Kaufmann Series in Data Management Systems
• Russell S , Norvig P - Artificial Intelligence - A
Modern Approach, Second Edition
How do we define
Machine Learning?
Formal Definition

Machine Learning is a branch

of artificial intelligence and computer
science that focuses on the using data
and algorithms to enable AI to enact the
way that humans learn gradually
improving its accuracy.
Book Definition
Machine learning investigates how computers can learn (or
improve their performance) based on data. A main research
area is for computer programs to learn to recognize complex
patterns and make intelligent decisions based on data.

For example, a typical machine learning problem is to

program a computer so that it can automatically recognize
handwritten postal codes on mail after learning from a set of
examples.
MNIST Numerical Character Dataset
Simple Analogy
A child to recognize different types of fruits. Instead of giving them a detailed
description of each fruit, you show them various examples.
1. Initial Exposure
• Present the child with pictures of apples, bananas, and oranges. Point out the colors,
shapes, and sizes.
2. Learning Through Examples
• Overtime sees many apples, bananas, and oranges. They start to notice patterns
apples are usually round and red or green, bananas are long and yellow, and
oranges are round and orange.
3. Testing Knowledge:
• A new fruit they haven’t seen before a pear and ask them what it is. Based on their
previous experiences, they might say it looks like an apple but is not quite the
same.
4. Refinement
• Correct them and explain why it’s different. The next time they see a pear, they’ll
remember the distinction.
Types of Machine Learning
Based on the type data we provide and type of algorithm employed while learning, there
are 3 broad categories

1. Supervised Learning
• The learning is done based on a labeled dataset - meaning that each training example is has a
corresponding output label. Ex: Classification, Regression

2. Unsupervised Learning
• Involves learning on data without labeled outputs. The algorithm/process/computer
tries to identify patterns and relationships in the data on its own. Ex: Clustering

3. Reinforcement Learning
• Learning is based on making decisions by taking actions in an environment to
maximize cumulative reward. The agent receives feedback in the form of rewards
or penalties based on its actions. Ex: A robot trying to map the environment based on
feedbacks.
Underfitting vs Overfitting
Based on the amount of the data the model has generalized or
learned or trained on, the learning level is defined in 2 broad
classes

1. Underfitting
• Occurs when a model is too simple to capture the underlying structure of the
data.

2. Overfitting
• Occurs when a model is too complex and captures not just the
underlying structure but also the noise or random fluctuations in the
training data.
Classification vs. Regression
Simple Regression
Clustering
 Sample ML Program
https://www.kaggle.com/code/hardikgupta0808/lab-2-simple-linear-regression
Questions & Answers
That’s all folks!
How do we define
Deep Learning?
Formal Definition

A machine learning algorithm is an

algorithm that is able to learn from data.
But what do we mean by learning? Mitchell
(1997) provides a succinct definition: “A
computer program is said to learn from
experience E with respect to someclass of
tasks T and performance measure P, if its
performance at tasks in T, as measured by
P, improves with experience E.”
Formal Definition

Deep learning is a specialized

subset of machine learning that
utilizes neural networks to process
data in a manner inspired by the
human brain.
Neural Network!?
Neural Network!?
Break it down to The Neuron.
The smallest Mathematical Unit of any neural network.
Break it down to The Neuron.
The smallest Mathematical Unit of any neural network.
House Price Prediction
Representation of Data!?
House Price Prediction using Neural
Network.
House Price Prediction using Neural
Network.
Backpropagation in Neural
Network.
Why it works
Stopping Criteria
Performance of The Neural
Network.
Back to Deep Learning
Utilizes artificial neural networks to process data in a manner inspired by the human
brain.

Deep learning is a type of machine learning that employs

multilayered neural networks, known as deep neural
networks, to automatically learn and extract features
from large amounts of data.

These networks consist of numerous layers of

interconnected neurons, allowing them to recognize
complex patterns in various forms of data, including images,
text, and audio.
Error Minimization (Gradient
Descent)
Error Minimization (Gradient
Descent)
Local Minima in Error Minimization
Confused about how?
- How an image is used in neural
network?
- How the text is used in neural
network for NLP?
- How the audio is used in neural
networks?
Resources
• Kevin Knight, Elaine Rich, B. Nair - Artificial
Intelligence - Tata McGraw-Hill Education Pvt. Ltd.
(2010)
• Jiawei Han, Jian Pei, Hanghang Tong - Data
Mining Concepts and Techniques - The Morgan
Kaufmann Series in Data Management Systems
• Russell S , Norvig P - Artificial Intelligence - A
Modern Approach, Second Edition