Data Modelling
1. Write down the Relationship Between AI, ML, and DL.
• Artificial Intelligence (AI) is the broadest concept, referring to machines designed to mimic human
intelligence, including reasoning, problem-solving, and decision-making.
• Machine Learning (ML) is a subset of AI that enables systems to learn and improve from data and experiences,
rather than being explicitly programmed.
• Deep Learning (DL) is a subset of ML that uses multi-layered neural networks to learn complex patterns in
large datasets, inspired by how the human brain works.
2. Comparison among : AI vs ML vs DL.
Aspect Artificial Intelligence Machine Learning (ML) Deep Learning (DL)
(AI)
Scope Broad field: any Subset of AI: systems that learn Subset of ML: uses neural
system mimicking from data networks for complex tasks
human intelligence
Learning Logic, rules, Learns from structured data using Learns from unstructured or
Method reasoning, and algorithms large datasets using many
human-like decisions neural network layers
Key Algorithms Rule-based, expert Regression, decision trees, Convolutional, recurrent,
systems, ML, DL clustering, and deep neural networks
supervised/unsupervised learning
Data Can work with Needs structured data Needs large amounts of
Requirements limited or structured unstructured data (images,
data text, audio)
Hardware Low to moderate Moderate High; GPUs and massive
Needs parallel processing
Examples Smart robots, chess Email filters, Netflix Face recognition, voice
engines, Google recommendations, spam detection assistants, self-driving cars
Maps AI
Objective Mimic human Automate learning and predictions Advanced pattern
intelligence and from data recognition and automation
behavior for high accuracy
Hierarchy AI is the parent field ML is a subset of AI DL is a subset of ML
• AI ⟶ ML ⟶ DL: Think of AI as a full toolbox, ML as a specific tool, and DL as a high-tech version of that tool,
used for the toughest jobs.
This framework helps visually and conceptually clarify how AI, ML, and DL relate and differ.
3. Write some Real-world Applications of Machine Learning (ML).
Some notable ML applications include:
• Healthcare: Disease diagnosis (cancer, pneumonia), predictive analytics, drug discovery, medical imaging
analysis.
• Autonomous Vehicles: Perception of surroundings, object detection, navigation, and decision-making in self-
driving cars.
• Retail: Demand forecasting, dynamic pricing, customer segmentation, self-checkout, and store layout
optimization.
• Natural Language Processing (NLP): Chatbots, speech recognition, language translation, sentiment analysis.
• Recommendation Systems: Personalized content suggestions in e-commerce, streaming services.
• Anomaly Detection: Fraud detection, network security, equipment monitoring.
• Robotics and Automation: Intelligent task handling in manufacturing, logistics.
• Energy and Traffic Control: Energy management and adaptive traffic signaling based on real-time data.
These examples show ML's wide impact across healthcare, finance, transportation, retail, and more.
4. What is Deep Learning (DL)? How It Works?
Deep Learning is a subset of machine learning that uses artificial neural networks with many layers (deep neural
networks) to automatically learn representations from large amounts of data. It models complex patterns by
mimicking the human brain's structure.
How it works:
• DL models consist of multiple layers of neurons: input, hidden (multiple), and output layers.
• Data is fed into the input layer.
• Each neuron processes information through weights and biases and a nonlinear activation function.
• Intermediate layers automatically extract hierarchical features (e.g., edges, textures, shapes in images).
• The output layer produces the final prediction or classification.
• The model trains using backpropagation and optimization algorithms (like gradient descent) to minimize
error by adjusting weights.
• Deep learning requires large datasets and powerful hardware (GPUs) for effective training.
DL excels especially in unstructured data like images, speech, and text, performing tasks such as image recognition,
natural language processing, and speech recognition with high accuracy.
5. Difference Between Machine Learning and Deep Learning ?
Feature Machine Learning (ML) Deep Learning (DL)
Algorithms that learn from data to make Subset of ML using multi-layer neural
Definition
predictions/decisions networks for complex tasks
Simpler models like decision trees, SVM, linear Complex, deep neural networks with multiple
Model Complexity
regression hidden layers
Data Requires very large amounts of data, often
Requires structured and often smaller datasets
Requirements unstructured
Feature
Often requires manual feature engineering Learns features automatically from raw data
Engineering
Typically requires long training times and
Training Time Generally faster
powerful hardware
Hardware Needs Moderate High (GPUs, TPUs for parallel processing)
Often considered a "black-box" due to
Interpretability Easier to interpret and understand
complexity
Spam filtering, credit scoring, recommendation Image and speech recognition, autonomous
Applications
systems vehicles, NLP
This table highlights that deep learning is a specialized, more resource-intensive method within the broader
spectrum of machine learning, designed for handling complex, large-scale problems.
6. What do you mean by Artificial Neural Network (ANN)?
Answer:
An Artificial Neural Network (ANN) is a computational model inspired by the structure and function of the human
brain. It is composed of interconnected nodes called artificial neurons, arranged in layers: an input layer, one or
more hidden layers, and an output layer. Each neuron receives inputs, processes them using weights and an
activation function, and passes the output to neurons in the next layer. The network learns by adjusting connection
weights through training methods like backpropagation to minimize prediction errors. ANNs are the foundation of
many machine learning and deep learning applications, capable of recognizing complex patterns and making
predictions in tasks such as image classification, natural language processing, and time series forecasting.
If you want to read more about ANN then I suggest you that read this article : More detailed Information about ANN
7. What are some real-world examples of Deep Learning (DL)?
Answer:
Deep Learning is widely applied in various domains, including:
• Image Recognition: Identifying objects, faces, or scenes in photos and videos (e.g., facial recognition
systems).
• Speech Recognition: Voice assistants like Siri and Alexa convert spoken words into text.
• Natural Language Processing (NLP): Language translation, sentiment analysis, chatbots, and text
summarization.
• Autonomous Vehicles: DL helps self-driving cars perceive the environment and make driving decisions.
• Healthcare: Diagnosing diseases from medical images such as X-rays or MRIs.
• Recommendation Systems: Providing personalized content in platforms like Netflix, YouTube, and Amazon.
• Robotics: Enabling robots to perform complex tasks by understanding sensory inputs.
8. What do you mean by data? What are the types of data, specifically training data and testing data?
Answer:
Data refers to the information collected for processing by machine learning or deep learning models. This
information can be numbers, text, images, audio, or any measurable input used for analysis and learning.
• Training Data: This is the dataset used to train the model. It includes inputs along with their correct outputs
(labels), allowing the model to learn patterns and relationships. The model adjusts its parameters based on
this data to minimize errors.
• Testing Data: After training, testing data is used to evaluate the model’s performance on new, unseen
examples. It helps assess how well the model generalizes to real-world data and checks for overfitting
(where the model performs well on training data but poorly on new data).
The distinction is crucial: training data shapes the model, while testing data validates its effectiveness.
9. What is Rule-based approach & Learn-based approach?
Rule-based approach refers to a system where decisions or actions are made based on a set of explicitly defined
rules, typically in the form of "if-then" statements created by human experts. These rules define how a system
should respond to specific inputs by applying fixed logical conditions without learning or adapting from new data.
This approach is deterministic, transparent, and works well in well-defined, static environments where domain
knowledge can be fully encoded into rules.
Learn-based approach (commonly seen in machine learning) involves systems that automatically learn patterns and
rules from data rather than relying on predefined instructions. These systems build models by analyzing examples,
updating their knowledge with new data, and improving performance over time. This approach is probabilistic,
adaptive, and suitable for complex or dynamic problems where explicit rule creation is difficult or impractical.
10. Difference between Rule-Based and Learn-Based Approach
Feature Rule-Based Approach Learn-Based Approach (Machine Learning)
Uses predefined, explicit rules created by Learns rules and patterns automatically from
Definition
humans data
Automatic, through algorithms analyzing
Rule Creation Manual, by domain experts
input data
Static; does not change unless rules are manually Dynamic; adapts and improves from new
Adaptability
updated data
Probabilistic outcomes based on learned
Decision Process Deterministic outcomes based on fixed logic
patterns
Data Requires large and diverse datasets for
Can work with limited and simple data
Requirement effective training
Complexity Handles complex, ambiguous, and non-linear
Limited to well-defined, simple scenarios
Handling patterns
Transparency High; easy to interpret and debug Often less interpretable ("black-box" models)
Requires retraining but scales better with
Maintenance Can be cumbersome as rules grow larger
data
Feature Rule-Based Approach Learn-Based Approach (Machine Learning)
Expert systems, rule-based chatbots, fraud Spam filtering, recommendation systems,
Examples
detection via fixed rules image recognition
Rule-based systems excel in environments where rules are clear and unchanging, while learn-based systems excel in
complex, data-rich environments where adaptation is essential.
11. What is Supervised Learning? Provide examples and explain where it is used.
Answer:
Supervised Learning is a type of machine learning where the model is trained on labeled data, meaning the input
data is paired with the correct output. The model learns to map inputs to outputs by minimizing the error between
its predictions and the actual labels during training.
Examples of Supervised Learning:
• Handwritten Digit Recognition: Using datasets like MNIST to classify images of digits (0-9).
• Medical Imaging: Detecting diseases such as cancer or pneumonia from labeled medical scans.
• Spam Detection: Classifying emails into spam or non-spam based on labeled examples.
• Credit Scoring: Predicting whether a new loan applicant is likely to repay based on historical data.
• Sentiment Analysis: Classifying customer reviews as positive or negative.
Where is it used?
Supervised learning is widely used in industries such as healthcare (diagnosis, prognosis), finance (fraud detection,
credit risk assessment), e-commerce (recommendation systems), natural language processing (speech recognition,
translation), and computer vision (image classification, object detection).
12. What is Unsupervised Learning? Provide examples and explain where it is used.
Answer:
Unsupervised Learning is a machine learning approach where the model learns patterns or structure from unlabeled
data, meaning the data does not have predefined labels or outcomes. The goal is to discover hidden patterns,
groupings, or data distributions.
Examples of Unsupervised Learning:
• Clustering: Grouping customers based on purchasing behavior without predefined categories.
• Dimensionality Reduction: Reducing features for data visualization or compression (e.g., PCA).
• Anomaly Detection: Identifying outliers in network traffic for cybersecurity.
Where is it used?
Unsupervised learning is used in customer segmentation for marketing, anomaly detection in fraud or security, gene
sequence analysis in bioinformatics, market basket analysis for recommendations, and data preprocessing tasks.
13. What is Reinforcement Learning? Provide examples and explain where it is used.
Answer:
Reinforcement Learning (RL) is a machine learning paradigm where an agent learns to make decisions by interacting
with an environment. The agent receives feedback in the form of rewards or penalties and aims to maximize
cumulative reward through trial and error.
Example of Reinforcement Learning:
• Game Playing: AlphaGo uses RL to learn strategies to win the game of Go.
• Robotics: Robots learn to perform tasks like grasping objects by trial and error.
• Self-Driving Cars: Learning how to navigate complex traffic environments safely.
Where is it used?
RL is used in gaming, autonomous vehicles, robotics control, resource management, and recommendation systems
where sequential decisions are critical and labeled data is not available upfront.
14. What is the Difference Between Supervised, Unsupervised, and Reinforcement Learning?
Feature Supervised Learning Unsupervised Learning Reinforcement Learning
Environment interaction, reward
Data Labeled data Unlabeled data
feedback
Discover hidden patterns or Learn optimal actions based on
Goal Predict output based on input
groupings rewards
Predicted labels or continuous Clusters, associations, or data Policy to maximize cumulative
Output
values structure rewards
Learning Error correction based on Pattern finding without Trial-and-error with rewards and
Process known labels supervision penalties
Email spam detection, disease Customer segmentation, anomaly Game AI, robotic control, self-
Examples
diagnosis detection driving cars
15. What is Classification and Regression? Give suitable examples.
Answer:
• Classification: A supervised learning task where the output is a discrete category or class label.
Example: Email spam detection (classify as spam or not spam).
• Regression: A supervised learning task where the output is a continuous numerical value.
Example: Predicting house prices based on features like size, location, and number of rooms.
16. What is Clustering and Association? Give suitable examples.
Answer:
• Clustering: An unsupervised learning method that groups similar data points together without predefined
labels.
Example: Grouping customers by purchase behavior for targeted marketing.
• Association: A rule-based learning method used to find relationships between variables in large datasets.
Example: Market basket analysis showing customers who buy bread often buy butter as well.
17. What is a Convolutional Neural Network (CNN)?
Answer:
A CNN is a specialized ANN designed for processing structured grid data like images. It uses convolutional layers with
filters to automatically detect spatial hierarchies of features such as edges, textures, and shapes. CNNs are widely
used in image recognition, video analysis, and computer vision tasks.
18. What is a Neural Network (NN)?
Answer:
A Neural Network is a general term for layers of interconnected nodes (neurons) that simulate a simplified brain-like
structure and learn patterns from input data through training. ANNs and CNNs are types of neural networks
specialized for different tasks.
19. What is the Working Process of a Neural Network?
Answer:
• Input Layer: Receives raw data and passes it forward.
• Hidden Layers: Multiple layers where neurons compute weighted sums of inputs, apply activation functions
to introduce non-linearity, and extract features.
• Output Layer: Produces the final output such as classification or regression results.
The network is trained iteratively by comparing outputs with actual labels, using backpropagation and optimization
algorithms (e.g., gradient descent) to adjust weights and minimize error.
20. How does AI make decisions? Explain Perceptron with an example.
Answer:
AI systems, especially neural networks, make decisions by processing inputs through layers of weighted connections
and applying activation functions. The Perceptron is a simple neural network unit that takes multiple inputs, applies
weights, sums them, and passes through an activation function to produce a binary output (e.g., yes/no decision).
Example: A Perceptron can decide whether an email is spam (output 1) or not spam (output 0) based on features
like presence of certain words by assigning weights to each input feature.
21. How does the Human Neural Network work? Give a suitable example.
Answer:
The human neural network consists of billions of neurons connected via synapses. Neurons receive electrical
impulses, process them, and transmit signals to other neurons. For example, when you touch a hot surface, sensory
neurons send signals to the brain, which processes the information and sends signals via motor neurons to withdraw
your hand quickly, demonstrating perception, processing, and response.
