Scribd
Upload
17 views42 pages

Lecture 1

The document is an introduction to neural networks, covering the history and applications of machine learning (ML), including supervised, unsupervised, and reinforcement learning. It highlights the advantages of deep learning, particularly in handling large datasets and complex problems like image classification and natural language processing. The course outline includes topics such as perceptrons, optimization methods, convolutional networks, and advanced topics like GPT and GAN.

Uploaded by

lokr.789
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
17 views42 pages

Lecture 1

The document is an introduction to neural networks, covering the history and applications of machine learning (ML), including supervised, unsupervised, and reinforcement learning. It highlights the advantages of deep learning, particularly in handling large datasets and complex problems like image classification and natural language processing. The course outline includes topics such as perceptrons, optimization methods, convolutional networks, and advanced topics like GPT and GAN.

Uploaded by

lokr.789
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/ 42

Introduction to Neural Netwrks

Prepared by: Dr / Doaa Gamal


Assistant professor at Faculty of Engineering, Suez Canal University
([email protected])
AI & ML & DL
2
AI & ML & DL
3
History of ML
4
History of ML
5

• 2020—present : “AI spring” highest levels of interest and funding


Machine learning
6

 For many problems, it's difficult to program the correct behavior by


hand
• recognizing people and objects
• understanding human speech
 system needs to adapt to a changing environment (e.g. spam
detection)
 want the system to perform better than the human programmers

 Machine learning approach: program an algorithm to

automatically learn from data, or from experience


ML workfow
7
Machine learning
8

Machine learning

Supervised learning Unsupervised learning Reinforcement learning


(classification, Regression) (clustering) (decision making)
Supervised learning (classification)
9

Discrete output

Teach by examples
Supervised learning (classification)
10

Discrete output
Supervised learning (classification)
11
Supervised learning (classification)
12

• Binary Classification problem: want correct


decision (play, no play)
• Input features X: (outlook, temp, humidity,
windy)
• Output classes Y: (play, no play)
• Training samples: number of rows (14)
• Required: function f which maps X to Y
Supervised learning (regression)
13

Continuous output

Teach by examples
Supervised learning (regression)
14
Unsupervised learning
15
Unsupervised learning (e.g. clustering)
16

Teach by examples
Unlabeled data
(unknown Y)
Unsupervised learning (e.g. clustering)
17
Reinforcement learning
18

Teach by experiment
APPLICATIONS
Computer vision and AI
20
Computer vision & AI (Facial detection and recognition)
21
Computer vision & AI(self-driving cars)
22
Computer vision & AI(medicine, biology and healthcare)
23
Computer vision & AI
24
Natural language processing
25
Sequential data
26
Health
27
Entertainments
28

Games (Atari- chess- Go- combat games)


Military
29
Telecommunications
30
Telecommunications
31

Optimal
transmitter
Performance receiver
optimization design

Channel
estimation
AI Applications in different fields
32
ML ALGORITHMS
Machine learning known algorithms
34

Machine learning

Supervised learning Unsupervised learning Reinforcement learning


(classification, Regression) (clustering) (decision making)

Linear/ Logistic regression K-means Deep Q-Net (DNN)


Deep Neural networks EM
Decision trees - Random forest Hierarchal clustering
Support vector machines (SVM) NN
K-Nearest Neighbor (KNN)
Why deep learning
35
ML Vs DL
36
ML Vs DL
37

 Deep Learning out perform other techniques if the data size is large.
 Deep Learning techniques need to have high end infrastructure to train in
reasonable time.
 When there is lack of domain understanding for feature introspection, Deep
Learning techniques outshines others as you have to worry less about feature
engineering.
 Deep Learning really shines when it comes to complex problems such as
image classification, natural language processing, and speech recognition.
Why now
38
Course outline
39

 Simple perceptron for classification, and Backpropagation


 Multilayer Perceptron for deep learning (FFNN),
 optimization methods for deep networks,
 Tricks of the Trade in deep learning
 Convolutional networks (CNN),
 Recurrent neural networks (RNN),
 Reinforcement learning: applications, Q-learning, Deep
Reinforcement learning.
 Introduction to Advanced topics (GPT, GAN,…)
Grade distribution
40

 Midterm 25%
 Attendance, Lab work and Project 25%
References
41

 MIT 6.S191,”Introduction to Deep Learning”


 Dive into Deep Learning (available online http://d2l.ai/)
 Deep Learning, Ian Goodfellow and Yoshua Bengio and Aaron
Courville (available online http://www.deeplearningbook.org/)
 • Neural Networks and Deep Learning, Michael Nielsen (available
online http://neuralnetworksanddeeplearning.com/)
 Online courses by Andrew Ng
 Lecture slides “prof Mahmoud Khalil, Ain shams university”
Thank You

You might also like