Scribd
Upload
4 views4 pages

Aih 2

The document outlines the definitions, advantages, disadvantages, workings, typical case studies, and examples of Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL), and Transfer Learning. AI focuses on mimicking human reasoning, ML automates decision-making through data, DL utilizes artificial networks for complex tasks, and Transfer Learning leverages pre-trained models for efficiency. Each technology has unique applications and challenges, such as bias in AI, data dependency in ML, computational costs in DL, and potential overfitting in Transfer Learning.

Uploaded by

Ke Dar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
4 views4 pages

Aih 2

The document outlines the definitions, advantages, disadvantages, workings, typical case studies, and examples of Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL), and Transfer Learning. AI focuses on mimicking human reasoning, ML automates decision-making through data, DL utilizes artificial networks for complex tasks, and Transfer Learning leverages pre-trained models for efficiency. Each technology has unique applications and challenges, such as bias in AI, data dependency in ML, computational costs in DL, and potential overfitting in Transfer Learning.

Uploaded by

Ke Dar
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 4

Difference Between AI, ML and DL, Transfer Learning.

Define,
Advantage, Disadvantages, Working, Typical Case Study, Example?

1. Artificial intelligence (AI)

Definition.

This involves considering the fundamentals of engineering and computer


science is a subfield that researches systems that can operate at the same
level as humans in functions like reasoning or decision making.

Advantages.

 Job automation.
 Increased productivity and work optimization.
 Better analysis of data results in better decisions.
 More intelligent services are rendered, such as Virtual assistant.

Disadvantages.

 Costly to develop and to maintain these systems.


 Lack of diverse development personnel leads to bias.
 Robots replace some workers.
 These tools can be misused imagine surveillance applications.

How it Works:

AI combines avast of data and computation and algorithms to achieve tasks


that require intelligent behavior. Representation of knowledge, logic
programming, and natural language processing are some of the
methodologies available in the market.

Typical Case Study:

An area concerned with AI is health assistive technology, chatbots can act as


a maintainence agent to patients, and computer can assist doctors to
diagnose patients suffering from depression accurate.

Example:

Virtual assistants include Siri, Alexa, Google assistant, etc.

2. Machine Learning (ML):

Definition.

This is one of AIs applicable technologies that does not require explicit
coding where the machine uses data to learn everything.
Advantages.

 ML removes human from complex decisions making it automated.


 It’s Smart and suggestable, e.g help in cross selling or upselling clients.
 Can manage significant data sets successfully.

Disadvantages:

 High volume of data must be acquired.


 Data is quality-dependent.
 Costly in terms of computation.

How it works:

Machine Learning uses supervised learning, unsupervised learning and


reinforcement learning algorithms. These algorithms take labeled data or
unlabeled data as input to devise a model that can recognize patterns and
make predictions.

Typical Case Study:

Detection of fraud in the banking sector with the use of past transaction
data.

Example:

Netflix’s recommendation engine.

3. Deep Learning (DL):

Definition:

Deep learning is a type of machine learning which is centered around


utilizing artificial networks to tackle and solve intricately complex challenges
that encompass images, text and audio.

Advantages:

 Possesses vast superiority over other approaches when dealing with


data that is unlabeled or unstructured such as images and audio.
 Technically, proficient in delivering unsurpassed or unparalleled results
across several fields.
 Has the effectiveness of learning by itself eliminating the need for
intervention during learning.

Disadvantages:

 Highly computationally expensive.


 Involves processes necessitating multitude of labeled datasets for the
machine to be trained.
 Due to the existing black box, this form is not interpretable.

How it Works:

MLPs (multi-layer perceptron) is combined with CNN to form the RNN wherein
the first layer trained classifies features from the images while the rest of the
layers extract features automatically. For images, they use CNNs while for
sequence, RNNs are employed.

Typical Case Study:

The rending and construction of an image is done through the use of panels
and layers and together they form an integrated image application. Self‐
driving vehicles are one of the case studies where DL was used, it utilized
deep learning in combination with multiple applications that were capable of
object detection and decision-making strategies.

Example:

The face recognition system embedded in smartphones is an example of a


device that utilizes Deep Learning for computer vision.

4. Transfer Learning

Definition:

Transfer Learning or transfer of knowledge is the process of using a pre-


trained model on one task to aid in a similar but different task thereby
reducing the training time and computing resources required.

Advantages:

 Lessens the need for data and computing resources.


 Takes less time to develop a model.
 Helps when the labeled data is not enough for some of the tasks.

Disadvantages:

 Tends to overfit if the source and target domain data sets are very
different.
 Has not worked well for tasks that are completely dissimilar.

How it Works:
The methodology involves starting with a model trained on huge data set say
ImageNet, where the lower layers of the model known as the feature
extractors are kept, and the upper layers are trained for the new task.

Typical Case Study:

A typical case involves text sentiment analysis with the help of BERT with
pre-trained weights.

Example:

Modifying a ResNet model which was previously trained on general tasks to


better suit medical imaging tasks.

You might also like