Page 201 - AI Computer 10

In Supervised learning, the model is trained using a labeled dataset where each input is paired with the
correct output. The goal is to learn a function that maps inputs to corresponding outputs.
The common examples of supervised learning applications include:

● Spam Detection: Email services use labelled datasets of emails, tagged as "spam" or "not spam," to
train models that filter incoming messages.
● Credit Scoring: Financial institutions use historical loan data labelled with outcomes (default or no
default) to predict the likelihood of repayment for new applicants.
Conversely, unsupervised learning involves training a model without labelled outputs, aiming to uncover
hidden patterns or structures within the input data. The common applications of unsupervised learning
include:
● Customer Segmentation: Companies such as Netflix use unsupervised learning to analyze customer
viewing habits and create clusters of similar users for tailored recommendations.
● Anomaly Detection: Industries such as banking employ unsupervised models to identify unusual
patterns in transaction data that may indicate fraudulent activities.

4. Discuss the role of deep learning in the advancement of artificial intelligence applications. How do neural
networks contribute to achieving complex tasks?

Ans. Deep learning has significantly transformed the artificial intelligence (AI) applications by providing
powerful tools for handling vast datasets and performing complex computations. It utilises neural
networks with many layers — hence the term "deep" — to model highly complex relationships in data.

Neural networks consist of interconnected layers of nodes (neurons) that work similarly to the human
brain. Each neuron processes input from previous layers and passes its output to subsequent layers.
The multi-layered architecture allows deep learning models to automatically extract relevant features
from data, minimising the need for manual feature engineering. This capability is particularly crucial
in applications involving high-dimensional data, where traditional machine learning algorithms may
struggle to identify patterns. For instance:

● Image Recognition: Systems like facial recognition software employ convolutional neural networks
(CNNs), a specialised type of neural network adept at processing grid-like data, such as images. They
can identify and classify the intricate details and patterns in visual content.

● Natural Language Processing (NLP): Recurrent neural networks (RNNs), another variant of neural
networks, are well-suited for tasks involving sequential data, such as language translation or sentiment
analysis. They can process context and semantics over varying input lengths, enabling applications like
chatbots and virtual assistants

5. Explain learning based approach in the context of email filters.
Ans. A learning-based spam email filter is a computer program that determines if an incoming email is spam
or not. Instead of being manually programmed with specific rules to identify spam, this filter learns from
examples of emails that have already been marked as spam or legitimate during a training phase.
In the training phase, the filter receives a large set of emails, each labeled as either spam or non-spam.
It looks at different aspects of these emails, like the words and phrases used, the address of the sender,
and whether these emails have attachments or not. By using machine learning algorithms, the filter
learns to recognise patterns that help it tell the difference between spam and legitimate emails.
After training, the filter can categorise new incoming emails as spam or not spam, based on the patterns
it has learned. This helps users manage their email inboxes more efficiently. It also adds new emails that

67
67