Scribd
Upload
16 views26 pages

1 - ML Intro 24

The document discusses the fundamentals of Machine Learning (ML) and its applications, highlighting its ability to learn from examples rather than being explicitly programmed. It covers various ML tasks such as image classification, natural language processing, and recommendation systems, along with notable successes in the field. The resurgence of ML is attributed to advancements in big data, hardware, and software, enabling more effective algorithms and applications.

Uploaded by

walidlabied2002
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
16 views26 pages

1 - ML Intro 24

The document discusses the fundamentals of Machine Learning (ML) and its applications, highlighting its ability to learn from examples rather than being explicitly programmed. It covers various ML tasks such as image classification, natural language processing, and recommendation systems, along with notable successes in the field. The resurgence of ML is attributed to advancements in big data, hardware, and software, enabling more effective algorithms and applications.

Uploaded by

walidlabied2002
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 26

Machine Learning

The beating heart of AI

Lecturer: Prof. Mounir Ghogho


Lab instructors: Safaa Berkani & Hamza Gamouh
A programming challenge:
Write a program to automatically identify handwritten digits

Application: identifying handwritten zip codes to sort mails


Another programming challenge
Another programming challenge
“Cat” or “Not Cat”
Another programming challenge
“Giraffe” or “Not Giraffe”

Do we know how our brain does it?


Traditional programming

• Writing a program that encodes a set of rules to solve a problem.


• Limitation: in many cases, it is untractable to specify these rules, e.g.
given a picture determine whether it contains a cat

Data
Computer Output
Program
Traditional Programming
Machine learning
o Learning system is not explicitly programmed to solve the problem; it
develops its own programs by extracting information from examples:
• Examples of how it should behave
• Trial-and-error experiences trying to solve the problem

Input data
Computer Program
Output data
Machine Learning

o A ML algorithm uses the examples (training set) to produce a program


that perfoms the task. If the job is done well, the program will give the
correct answers for (most of) the new cases.
A ML task: image classification

Input data

Output data (labels)


When do we use Machine Learning?
• ML is used when:
• Human expertise does not exist
• Humans can’t fully explain their expertise (e.g. speech recognition,
computer vision)
• Models must be customized (e.g. personalized medicine)
• Models are based on huge amounts of data (e.g. genomics)

• Machine Learning is not always useful:


• There is no need to “learn” to calculate payroll
• There is no need to “learn” to plot a function
Machine Learning: formal definitions
An early success of Machine Learning
Remarkable surge in ML performance
from 2012
Neural Networks date back decades, so why the resurgence?
Stochastic Gradient
1952
Descent

Perceptron 1. Big Data 2. Hardware 3. Software


1958
• Learnable Weights
• Larger Datasets • Graphics • Improved
• Easier Collection Processing Units Techniques
& Storage (GPUs) • New Models
1986 Backpropagation • Massively • Toolboxes
• Multi-Layer Perceptron Parallelizable
1995 Deep Convolutional NN
• Digit Recognition

© MIT 6.S191: Introduction to Deep Learning


A ML task: image classification
ImageNet Large-Scale Visual Recognition Challenge (ILSVRC)

Every year since 2010, ILSVRC


uses a subset of ImageNet with
roughly 1000 images in each of
1000 categories.

ImageNet is a dataset of over 15 million labeled high-resolution images belonging to roughly 22,000 categories.
The images were collected from the web and labeled by human labelers using Amazon’s Mechanical Turk crowd-
sourcing tool (https://www.image-net.org).
A ML task: image classification
ImageNet Large-Scale Visual Recognition Challenge (ILSVRC)
Other applications of ML
Classification
• Example: Spam detection • Example: Fraud detection
Other applications of ML
Classification
• Example: sentiment analysis

Natural language processing (NLP)


Other applications of ML
Object detection (localization & classification)

Computer vision

Diagnosing diabetic retinopathy Lung cancer detection using X-Rays Bladder cancer detection during cystoscopy
Other applications of ML
Computer vision for self-driving cars
Other applications of ML
Recognizing patterns
Speech recognition Face recognition Licence plate identification
Other applications of ML
Recommendation system
Netflix film recommendations
Other applications of ML
Recommendation system
Other applications of ML
Machine Translation

Natural language processing (NLP)


Recent popular successes of ML Chat-GPT
(2022)
AlphaGo - Google
(2017) Text/image
/video
IBM Project debater
Go game generation
(2014)
IBM Watson
(2011) Debating
IBM Deep Blue Understanding
(1997) & answering
Chess game questions
Generative AI: ChatGPT
Prompt A function with billions of parameters

xxx xxx xxx xxx xxx


Probabilistic choice Large Langage Model (autoregressive)
xxx xxx xxx xxx xxx xxx

xxx xxx xxx xxx xxx xxx xxx

xxx xxx xxx xxx xxx xxx xxx xxx

xxx xxx xxx xxx xxx xxx xxx xxx xxx

xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx self-supervised learning reinforcement learning with
human feedback
Training datasets
Output
Types of ML
End

You might also like