Answers natural language
(iv) (c) To understand
1. () (6) Icons (v) (c) Instance
Segmentation
(i) () Inproved mood and relaxation (vi) (c) Natural Language
Processing
events
(ii) (c) Rising sea levels and extreme weather demands, financial
6. Sarah's stress stems from job personal time. To address
(iv) (a) Exclamatory concerns, and inadequate
(v) (a) Both A and Rare correct and R is the correct tasks, seek financial
these, she could prioritize
explanation of A advice, and schedule regular breaks for relaxation
balance her
(vi) (b) Consistently meeting deadlines and and personal activities, helping to
maintaining high standards responsibilities and reduce stress.
2. (i)) (a) Both A and R are correct and R is the correct 7. Descriptive feedback provides
clear, actionable
explanation of A learner's specific
suggestions tailored to the
(i) (a) Training performance, rather than just a general evaluation. It
for growth,
A, Mistake Alert helps identify strengths and areas
Many students confuse Optimization with Training. guiding the learner towards more effective strategies
Optimization is part of training where weights are and practices.
sustainable
adjusted using techniques like gradient descent. But 8. The primary objective of the
Training refers to the entire process - including development policy is to create a balanced approach
feeding data, calculating loss, and adjusting weights Mn that addresses environmental, economic, and social
-making it the correct broader term. needs. By reducing carbon emissions and promoting
(iii) (a) Both Statement 1 and Statement 2 are true renewable energy, the policy aims to ensure
(iv) (b) Object detection long-term ecological health, economic stability, and
(v) (b) Stop words iw ob improved quality of life for residents.
(vi) () To assess how wellthe chosen model will work 9. Multitasking OS allows multiple tasks or processes
in future to run simultaneously by managing CPUtime
among them,whereas multithreading OS involves
3. ) () Establishing relationships between preceding multiple threads within a single process running
and succeeding sentences concurrently. Multitasking handles different
(i) (b) Deep Neural Network applications or processes at once while
(ii) () Asigning an input image one label from a multithreading improves efficiency and
fixed set of categories performance for tasks within the same application.
(iv) Google Assistant 10. Open-mindedness allows entrepreneurs to explore
(v) (b) F1 Score new ideas, adapt to changes, and embrace
(vi) () By exploring actions and receiving feedback in innovative solutions. It fosters creativity and helps
the form of rewards or penalties in understanding diverse perspectives, which can
lead to better decision-making and problem-solving
4. ) (6) Recognizing patterns in complex data like ultimately driving business growth and success.
images and text
(ii) (a) Both Statement 1 and Statement 2 are true 11. Image segmentation divides an image into parts or
(i)(c) Recall regions to isolate objects or areas of interest, making
it easier for computers to analyze specific
Note components separately.
Know that Recall =TP / (TP + FN). It measures how 12. Lexical analysis in NLP is the process of breaking
well a model captures all actual positive cases, down text into its smallest meaningful units
making it essential in medical or fraud detection (tokens), like words and phrases,
tasks where missing positives is costly. to prepare itIO
further processing.
(iv) (b) To group data points based on similarity 13. Classification accuracy is the
predictions made by the model. percentage or cO
(v) (c) Spotify However, it is not
(vi) (b) To understand the aim and objective of the always reliable, especially for imbalanced datasets.
project In such cases, metrics like precision, recall, and
F1-score provide a better evaluation of model
5 () (a) Both Statement 1 and Statement 2 are true performance.
(ii) (b) first train it using the data.
14. Real-life examples, such as AI in
(ii) (a) Complexity of human languages agriculture, or transportation, helphealthcare,
connect
iheoretical concepts to practical SAMPLE QUESTION PAPER 06 93
demonstrate how machines
can
benefits, They
diagnosing diseases, assist humans in 19. Step 1Tokenization
monitoring
navigating roads with minimal erops, or Docl al, is, transforming, the, world
15. CClustering model/Unsupervised
human input. Doc2 ai, and, ml, are, changing, industries
organise thee unlabelled
input learning is used to Step 2Vocabulary
based on features. data into groups [al, is, transforming, the, world, and, ml, are,
Clustering is an changing, industries)
which can clusterunsupervised
unknown learning algorithm Step 3Vectorization
patterns or trends identified data according to the
out
observed might be the ones whichof it. The patterns Term Doc1 Doc2
developer or it might even are known to the af 1 1
unique patterns outof it. come up with some is 1
16. Supervised Learning In transforming
supervised learning the
model is trained with labeled the 1
(e.g, email spam filtering). data to predict outputs world 1 0
Unsupervised
model finds Learning In unsupervised learning the and 1
hidden patterns m 0
unlabeled data (e.g., customer or structures in
Reinforcement Learning In thissegmentation).
are 1
the agent learns by
receiving rewards or penalties based changing 0 1
in an environment (e.g., Al on actions taken industries 1
playing a game).
17. Machine Learning (ML) is a subset 20. The three major domains of Artificial Intelligence
of
Intelligence where machines are trainedArtificial
to learn
are:
from data and make decisions or 1. Data Science (Statistical Data and Analysis)
predictions
without being explicitly programmed. This domain focuses on understanding and
ML models
typically require feature extraction by humans and interpreting large volumes of data using
perform well on structured data. statistical methods.
Example Aspam email filter that learns from email Example Predicting customer buying patterns in
features (like keywords or sender) to classify emails e-commerce platforms.
as spam or not. 2. Computer Vision It enables machines to interpret
and analyze visual information from the world
Deep Learning (DL) is aspecialized branch of such as images or videos.
Machine Learning that uses artificial neural
networks with many layers (hence "deep"). It Example Face recognition systems used in
automatically learns features from unstructured smartphones for unlocking.
data like images, text, or audio. 3. Natural Language Processing (NLP) NLP
helps
Example An image recognition system that detects computers understand, interpret, and respond to
and classifies objects in pictures without manual human language (text or speech).
feature extraction. Vheonbnd Example Chatbots used in customer support or
Coen alda Virtual assistants like Alexa and Siri.
Deep Learning generally requires more data and 21. The confusion matrix would be:
computing power than traditional Machine
Learning but offers higher accuracy in complex Predicted Predicted
tasks. arlt ni colbbid alPositive
18, Image resolution refers to the number of pixels in a n hActual Positive Negative
os0 (TP 15 (FN)
Image, and pixel values represent the brightness or la nActual Negativeoll 10 (FP)
resolution images have 80 (TN)
color of each pixel. Higher
In this matrix:
more detail, enabling the model to detect finer
features. "True Positives (TP) The model correctly predicts 50
tasks l i k e g positive cases.
Accurate pixel values are crucial for
teature extraction andobject detection. If the "True Negatives (TN)The model correctly predicts
may miss small or 80 negative cases.
tesolution is too low, the model resolution
accuracy. Proper " False Positives (FP) The model incorrectly predicts
subtle patterns, reducing ensure the CV model 10negative cases as positive.
dnd consistent pixel data
classification and
Pertorms better across tasks like llov False Negatives EN) The model incorectly
Segmentation. predicts 15 positive cases as negative.