(A) CNN and ANN:

CNN – Convolutional Neural Network: CNN is a Deep Learning algorithm that takes an input image, assigns importance (learnable weights and biases) to various aspects/objects in the image, and differentiates one from the other. It consists of layers: Convolution Layer, ReLU, Pooling Layer, and Fully Connected Layer.

ANN – Artificial Neural Network: ANN is a computational model inspired by the human brain, consisting of interconnected nodes (neurons) organised in layers that process information to learn patterns and make decisions.

(B) Identification of Layers:

• Box 1 → Hidden Layer (the intermediate processing layer between input and output)
• Box 2 → Output Layer (produces the final prediction/result)

Source: Chapter 5, Section 5.6 Convolutional Neural Networks (CNN)

---

• For (A), the examiner expects the full form + a clear definition for each term. CNN's definition is directly from the textbook (Section 5.6); ANN is a prerequisite concept — keep it concise but accurate.
• For (B), in a standard feedforward neural network diagram, the layers in order are: Input Layer → Hidden Layer(s) → Output Layer. Since Box 1 and Box 2 are the unlabelled middle and last layers respectively, they correspond to Hidden Layer and Output Layer. Always name them precisely — partial credit may be lost for vague answers.
• Do not expand on CNN layers in sub-question (A) beyond what is necessary; save word count as this is only 3 marks.

Differentiation:

| Basis | Artificial Intelligence (AI) | Machine Learning (ML) | Deep Learning (DL) |
|-------|-----------------------------|-----------------------|--------------------|
| Definition | Technique that enables computers to mimic human intelligence | Enables machines to improve at tasks with experience | Enables software to train itself using vast amounts of data |
| Scope | Broadest – umbrella term | Subset of AI | Subset of ML |
| Data needed | Works on algorithms and data fed to it | Learns from new data in each iteration | Requires huge/vast amounts of data |
| Algorithm | Predefined | Machine learns its own rules | Uses multiple ML algorithms (e.g., ANN) |

Venn Diagram:

```
┌─────────────────────────────┐
│ Artificial Intelligence │
│ ┌─────────────────────┐ │
│ │ Machine Learning │ │
│ │ ┌─────────────┐ │ │
│ │ │ Deep │ │ │
│ │ │ Learning │ │ │
│ │ └─────────────┘ │ │
│ └─────────────────────┘ │
└─────────────────────────────┘
```

• AI is the outermost/umbrella concept.
• ML is a subset of AI.
• DL is a subset of ML (uses multiple ML algorithms working together).

Source: Chapter 2, Section 2.1 – Revisiting AI, ML, DL

---

• Examiners expect a clear table for differentiation (3 rows of differences = ~2 marks) and a labelled nested Venn diagram (~2 marks). Both parts are compulsory.
• Key phrase to remember: AI ⊃ ML ⊃ DL (nested/funnel relationship).
• The diagram must be labelled with all three terms clearly shown as concentric/nested shapes — marks are awarded for labels.
• Do not mix up DL and ML definitions: DL uses vast data + multiple ML algorithms, while ML learns from experience/feedback.

| Basis | Supervised Learning | Unsupervised Learning |
|---|---|---|
| Data | Uses labeled data for training. | Uses unlabeled data for training. |
| Goal | Predicts a specific output (classification/regression). | Discovers hidden patterns or clusters in data. |

Example: Spam email filter (Supervised) vs. Customer grouping/clustering (Unsupervised).

Source: Chapter 2, Section 2.2 – Modelling

---

• Examiners look for the two clear differences — the most scoring points are labeled vs. unlabeled data and predict output vs. discover patterns.
• Presenting differences in a table format is preferred for "differentiate" questions — it earns full marks neatly.
• Adding a one-line example for each is optional but shows understanding and can secure boundary marks.
• Avoid writing long paragraphs; two crisp rows in a table are sufficient for 2 marks.

Two characteristics of a Classification Model are:

1. Works on discrete/categorical data — The output is a category or class label (e.g., "spam" or "not spam", "hot" or "cold"). The predicted values are not continuous.

1. Uses labeled data for training — It is a type of supervised learning where the training dataset contains predefined class labels to which new data points are assigned.

Source: Chapter 2, Classification Model (Sub-categories of Supervised Learning)

---

Examiners expect students to clearly name each characteristic and briefly explain it. The two most important points from the textbook are: (i) classification works on discrete (not continuous) data, and (ii) it assigns data to predefined class labels using labeled training data. Avoid confusing classification with regression (which predicts continuous values) or clustering (which has no predefined labels).

(C) Predicting tomorrow's temperature.

Regression models predict a continuous value (e.g., temperature, price). Predicting tomorrow's temperature outputs a numerical value, making it a regression problem.

Source: Regression Model, chapter 2

---

• Why not (A)? Spam/not-spam is a discrete (categorical) output → Classification.
• Why not (B)? Grouping customers by behaviour with no predefined labels → Clustering (Unsupervised).
• Why not (D)? Recognizing faces assigns a label/category → Classification.
• The textbook explicitly states: "Temperature is a continuous variable… Regression models are well-suited for predicting continuous outputs." Always link regression to continuous numerical outputs.

(C) Association model — This recommendation ("customers who bought X also bought Y") identifies relationships/associations between products purchased together, which is exactly what an Association model does.

The Association model (a sub-category of Unsupervised Learning) finds interesting relationships between variables in a database — e.g., bread buyers also buy butter. The textbook explicitly states such associations are used to recommend items to customers. Classification predicts categories, Regression predicts continuous values, and Clustering groups similar data points — none of these match "co-purchase" pattern finding.

(D) Reinforcement Learning — Parking a car involves trial-and-error decision-making where the AI learns by receiving rewards for correct actions and penalties for mistakes, which is the core of Reinforcement Learning.

Source: Chapter 2, Section 2.2 – Reinforcement Learning

The textbook explicitly lists "Parking a car" as an example of Reinforcement Learning and states it "enables the computer to make a series of decisions that maximize a reward metric without human intervention." Supervised learning needs labelled data; unsupervised finds patterns; transfer learning is not covered as a main ML category here. Reinforcement Learning fits best because parking requires sequential decisions optimised through reward/penalty feedback.

(a) Clustering (2 marks)

Clustering is an unsupervised learning model that groups unlabeled data into clusters based on similarities or patterns identified by the machine.

• The model finds similar data points and places them in the same cluster, differentiating them from objects in other clusters.
• No predefined class labels are given; the machine discovers groups on its own.
• Example: A supermarket customer database is clustered into "grocery shoppers" and "non-grocery shoppers" without any prior labels. OTT platforms like Netflix use clustering for recommendations.

(b) Dimensionality Reduction (2 marks)

Dimensionality Reduction is an unsupervised learning technique used to reduce the number of features (dimensions) in a dataset while retaining important patterns and relationships.

• It helps simplify large, complex datasets, making them easier to process and visualize.
• It removes redundant or irrelevant features, improving model efficiency.
• Example: Compressing pixel data of images while preserving key features for pattern recognition.

Source: Chapter 2, Section 2.2 – Sub-categories of Unsupervised Learning Model

---

• Clustering is directly covered in the source passages with examples — always mention it is unsupervised and uses unlabeled data. Examiners expect the distinction from classification (no predefined labels).
• Dimensionality Reduction is listed as a sub-category of unsupervised learning models but not elaborated in the passages — give a definition + purpose + example for full marks.
• The key examiner expectation: link both to unsupervised learning and unlabeled data.

(C) It receives and passes the raw input data to the hidden layers.

The input layer is the first layer of a neural network. Its job is to acquire data and feed it to the network; no processing occurs at the input layer.

Source: Chapter 2, Section 2.3 Artificial Neural Networks

---

The passage explicitly states: "The job of an input layer is to acquire data and feed it to the Neural Network. No processing occurs at the input layer." This rules out options (A) and (D). Option (B) describes the output layer's role, not the input layer. Examiners expect you to recall that the input layer only receives and forwards raw data — all computation happens in the hidden layers.

True. ML and DL are both subsets of AI, but DL is a specific subset of ML (not all ML is DL), as DL uses multiple ML algorithms together to train on vast data.

Source: Chapter 2, Section 2.1 – Revisiting AI, ML, DL

---

The Venn diagram in the textbook shows AI as the outermost circle, ML inside it, and DL as the innermost circle within ML. This confirms: all DL is ML, but not all ML is DL. Examiners expect you to state "True" and briefly justify using the subset relationship.

(B) Artificial Intelligence

The passage directly states: "Artificial Intelligence, or AI for short, refers to any technique that enables computers to mimic human intelligence." This is the definition examiners expect you to recall. Source: Chapter 2, Section 2.1 – Revisiting AI, ML, DL.

Concept: Neural Network (Deep Learning)

The diagram illustrates a Neural Network, a Deep Learning model inspired by the human brain's structure.

Working:

• A neural network consists of interconnected nodes (neurons) organized in three layers:

1. Input Layer – Receives raw data (e.g., pixel values of an image).
2. Hidden Layer(s) – Multiple ML algorithms work together to automatically extract features and identify patterns from the data.
3. Output Layer – Produces the final result (e.g., identifying the object as a bird).

• Each node acts as a machine learning algorithm, processing information passed from the previous layer.
• The network is trained on large datasets, enabling it to develop its own algorithms and improve accuracy over time without explicit programming.

Source: Chapter 2, Section 2.1 & Sub-Categories of Deep Learning

---

• The diagram showing "organized machine learning algorithms performing tasks" points to Deep Learning / Artificial Neural Network (ANN) — multiple ML algorithms working together in layers.
• Examiners expect: name of concept, inspiration (human brain), three layers with their roles, and the key feature (automatic feature extraction from large data).
• Don't confuse ANN with simple ML — emphasize multiple algorithms and layers as the distinguishing features.
• 4 marks = concept identification (1) + three layers explained (2) + key characteristic (1).

(D) Musical intelligence describes a person's ability to recognize and create sounds, rhythms and sound patterns.

This is a direct factual recall question. Option (D) is correct as it matches the standard definition of musical intelligence (one of Howard Gardner's multiple intelligences). The source passages focus on AI/ML concepts and do not define musical intelligence, but among the options, (D) is the only one that correctly describes musical intelligence — the others describe logical-mathematical (A), linguistic (B), and naturalist (C) intelligences respectively.

(B) Classification

A diagram showing data points sorted into predefined labeled categories represents a Classification model, which is a supervised learning approach that assigns inputs to discrete classes.

• Classification uses labeled data and assigns outputs to distinct categories (e.g., spam/not spam, hot/cold). The diagram typically shows data divided into separate labeled groups/categories.
• Clustering (unsupervised) finds groups without predefined labels — eliminate if the diagram shows labeled categories.
• Regression predicts continuous values — eliminate if the diagram doesn't show a trend line or continuous output.
• Key examiner tip: Look at whether the groups/outputs are labeled and predefined (→ Classification) or discovered by the model (→ Clustering).

(B) first train it using the data.

Before an AI model can predict any output, it must be trained on data so it can learn patterns and relationships from examples.

The passage clearly states that under the learning-based approach, the AI model "gets trained on the data fed to it and then is able to design a model." Training always precedes testing/prediction. Testing (option A) comes after training to check accuracy — so "both A and B" is incorrect because training must happen first. The correct sequence is: train → then predict/test.

(C) (A) is correct but (R) is not correct.

Assertion is correct: AI refers to any technique that enables computers to mimic human intelligence. However, a fully automatic washing machine only follows pre-set programmed instructions and does not mimic human intelligence, so it is not artificially intelligent.

• The textbook defines AI as techniques that enable computers to mimic human intelligence — so Assertion (A) is true.
• A fully automatic washing machine operates on fixed, pre-programmed rules (it does not learn, adapt, or mimic human traits), so Reason (R) is false.
• Examiners expect students to distinguish between automated/programmable machines and genuinely AI-enabled machines. This is a standard trap in assertion-reason questions on this topic.

The two graphs represent the two sub-types of Supervised Learning models: Classification and Regression.

Figure 1 – Classification Model:
In this model, data is classified into discrete categories/labels. The straight dashed line acts as a decision boundary separating two groups. It works on discrete (non-continuous) data.
Example: Classifying emails as "spam" or "not spam" based on labeled training data.

Figure 2 – Regression Model:
In this model, a continuous curved/straight line is fitted through the data points to predict continuous output values. It works on continuous data.
Example: Predicting the price of a house based on features like number of bedrooms, carpet size, and garage area.

Both models learn from labeled data, which is the defining feature of Supervised Learning.

Source: Chapter 2, Section 2.2 – Sub-categories of Supervised Learning Model

---

• The question asks you to identify (name) and explain each model with examples — so you must do all three parts for each figure.
• Figure 1 shows a linear decision boundary → Classification (discrete output).
• Figure 2 shows a curve fitted to points → Regression (continuous output).
• Key distinguisher: discrete output = Classification; continuous output = Regression.
• Always mention that both fall under Supervised Learning (labeled data) — examiners often award a mark for this link.

Neural Networks:
Neural networks are loosely modelled after how neurons in the human brain behave. They can extract data features automatically without needing input from the programmer. A neural network is a system of organising machine learning algorithms to perform certain tasks. It is fast and efficient for solving problems with very large datasets, such as images.

Layers of a Neural Network:

1. Input Layer: The first layer. It acquires data and feeds it to the network. No processing occurs here.

1. Hidden Layer(s): These layers perform all the processing using weights and biases on each node. Each node runs its own machine learning algorithm on the received data. There can be multiple hidden layers depending on the complexity of the task.

1. Output Layer: The last layer. It receives the final processed data from the hidden layers and presents it to the user. Like the input layer, no processing occurs here.

Source: Chapter 2, Section 2.3 – Artificial Neural Networks

---

• The question is worth 4 marks, so examiners expect: a brief definition (1 mark) + explanation of all three layers with their roles (3 marks — 1 each).
• Key phrase to remember: "no processing at input/output layer" — this is a commonly tested point.
• Mention weights and biases for the hidden layer to show accurate understanding.
• Do not confuse "multiple hidden layers" with separate marks — mention it as a feature of the hidden layer briefly.

Recommended approach: Supervised Learning

For predicting disease outbreaks and allocating resources, supervised learning is preferred. The organisation has labeled medical records (patient diagnosis, outcomes, disease history), which serve as training data. The model learns patterns from this labeled data to make predictions — a core feature of supervised learning.

Unsupervised learning would be used only if no labels existed and the goal was merely to discover unknown patterns.

Source: Chapter 2, Section 2.2 – Supervised Learning

---

• Examiners expect you to name the approach and justify using the key concept (labeled vs. unlabeled data).
• The trigger words in this question are "predicting" (a supervised task) and "medical records" (existing labeled data — diagnoses, outcomes).
• Mentioning why the other approach (unsupervised) is not suitable adds a contrast point and strengthens the answer within the word limit.
• Do not over-explain; two focused points are enough for 2 marks.

| Basis | Machine Learning (ML) | Deep Learning (DL) |
|---|---|---|
| Definition | Enables machines to improve at tasks with experience; learns from mistakes. | Enables software to train itself using vast amounts of data with multiple ML algorithms. |
| Data Requirement | Works with relatively smaller datasets. | Requires huge/vast amounts of data. |
| Algorithm | Uses single ML algorithms; cannot develop its own. | Uses multiple ML algorithms (ANN/CNN); machines develop algorithms for themselves. |
| Subset | Subset of AI. | Subset of ML (more specific). |

Source: Chapter 2, Section 2.1 – Revisiting AI, ML, DL

---

• Examiners expect a tabular comparison for "differentiate" questions — it earns full marks cleanly.
• Key contrast points: DL needs vast data, uses multiple ML algorithms working together, and is a subset of ML, whereas ML is a subset of AI.
• Remember the hierarchy: AI ⊃ ML ⊃ DL (funnel/Venn diagram concept).
• Avoid writing lengthy paragraphs for a 2-mark differentiation — 2–3 clear points in a table are sufficient.

(c) Artificial Intelligence

The passage defines AI as "any technique that enables computers to mimic human intelligence," which includes making decisions, predicting, learning, and improving — exactly what the question describes. The other options are narrower concepts that fall within AI, not the umbrella term.

(d) A music streaming platform that suggests songs and playlists based on user listening history.

A recommendation system suggests items to users based on their past behaviour or preferences. Option (d) directly matches this — the platform analyses listening history to suggest relevant songs. Options (a), (b), and (c) describe a product catalogue, a search engine, and a review display system respectively, none of which involve personalised recommendations. The textbook also cites Spotify/Netflix-style platforms as classic examples of recommendation systems using clustering/unsupervised learning.

(a) Clustering

The diagram shows two groups of data points enclosed in separate ovals, representing data grouped by similarity without predefined labels — which is clustering, a form of unsupervised learning.

Clustering is a subcategory of unsupervised learning where the model groups similar data points together based on patterns, without any predefined class labels. The diagram clearly shows two distinct clusters (ovals), making option (a) the correct answer. Regression predicts continuous values (supervised), classification uses labeled categories (supervised), and reinforcement learning uses rewards/penalties — none of which match the diagram.

(d) A digital alarm clock that rings at a set time every morning.

A digital alarm clock follows a fixed, pre-programmed instruction (ring at a set time) — it does not learn, adapt, or make decisions. The other options (robotic vacuum, chatbot, facial recognition) all involve AI techniques such as navigation algorithms, natural language processing, or machine learning. AI refers to machines mimicking human intelligence through algorithms and data; a simple alarm clock merely executes a hardcoded rule with no intelligence involved.

Yes, it is valid to say that not all "smart" devices are AI-enabled.

"Smart" simply means a device can connect to the internet or perform automated tasks. AI, however, requires the ability to learn from data, make decisions, and improve over time. A device can be smart without having any AI capability.

Two commonly misunderstood examples:

1. Smart TV – A Smart TV connects to the internet and runs apps like Netflix or YouTube. It does not learn or make decisions on its own; it simply executes programmed functions. Hence, it is smart but not AI-enabled.

1. Smart Street Lights – These lights turn on/off automatically based on a timer or light sensor. This is pre-programmed automation, not AI, as no learning or data-based decision-making is involved.

---

• The question tests whether you understand the difference between "smart" (connected/automated) and "AI-enabled" (learns from data, makes decisions).
• CBSE examiners expect: one clear justification statement + two distinct, well-explained examples. Avoid listing examples without explaining why they are not AI.
• Do not confuse automation/connectivity with machine learning or intelligence — that distinction is the core of the answer.
• Keep each example to 2 lines: name it, describe what it does, state why it is NOT AI.

No, AI, ML, and DL are not the same — they are related but distinct terms:

• Artificial Intelligence (AI): Any technique that enables computers to mimic human intelligence. It is the broadest, umbrella term.
• Machine Learning (ML): A subset of AI. Enables machines to improve at tasks with experience by learning from new data and its own mistakes.
• Deep Learning (DL): A subset of ML. Enables software to train itself using vast amounts of data through multiple ML algorithms (e.g., ANN) working together. It is the most specific and advanced of the three.

Relationship: AI ⊃ ML ⊃ DL (funnel approach — many AI applications, fewer ML, fewest DL).

Well-labelled Diagram (Venn/Nested):

```
___________________________
| AI |
| _____________________ |
| | ML | |
| | ______________ | |
| | | DL | | |
| | |______________| | |
| |_____________________| |
|___________________________|
```

Source: Chapter 2, Section 2.1 — Revisiting AI, ML, DL

---

• Examiners expect a clear distinction of all three terms (definition + relationship), not just definitions alone.
• The diagram must be nested/Venn-type showing DL inside ML, ML inside AI — labelling all three circles is mandatory for full marks.
• The key phrase to remember: "funnel approach" — AI is widest, DL is narrowest.
• Do not say they are completely different; the answer must convey they are related but not interchangeable.

Dimensionality Reduction is a technique used in unsupervised learning to reduce the number of features (dimensions) in a dataset while retaining the most important information. It helps simplify complex data, remove irrelevant or redundant features, and make the data easier to process and visualize, improving the efficiency of machine learning models.

The source passages do not explicitly define Dimensionality Reduction — it is a standard unsupervised learning concept from the broader CBSE AI syllabus not covered in the provided extracts. Examiners expect students to mention: (1) it reduces the number of features/variables, (2) it retains important information, and (3) it is used to simplify data and improve model performance. Avoid writing lengthy paragraphs — two to three crisp sentences are sufficient for 2 marks.

(c) Clustering

Clustering is an unsupervised learning algorithm that groups unlabeled/unknown data based on patterns or similarities the model identifies on its own — no predefined labels are used. Regression and Classification are supervised learning types; Dimensionality Reduction is a different unsupervised technique.

The model is Supervised Learning. In this model, the dataset fed to the machine is labelled, meaning the data has tags/information attached to it used for training.

Source: Chapter 2, Section 2.2 – Supervised Learning

The key phrase in the question is "labelled data." Examiners expect you to directly name Supervised Learning and briefly justify it by linking labelled data to this model. Do not confuse it with Unsupervised Learning (unlabelled data) or Reinforcement Learning (reward-based). Just one line is sufficient for 1 mark.

(c) They are modelled on human brains and nervous system.

The passage states: "Neural networks are loosely modelled after how neurons in the human brain behave." Options (a), (b), and (d) are incorrect — neural networks work best with large datasets and extract features without human intervention.

False.

Regression is a type of supervised learning, but it works on continuous data (e.g., predicting salary or house price). It is the Classification model where data is classified according to labels and data need not be continuous.

The statement mixes up two sub-types of Supervised Learning. The textbook clearly states: Classification — data is classified according to labels, works on discrete (non-continuous) data; Regression — works on continuous data to predict a value. The statement incorrectly attributes classification's features to regression, making it False.