SlideShare a Scribd company logo
1 of 38
Download to read offline
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
2
Source: https://www.youtube.com/watch?v=eWSbIXSbMis
AI vs AI
Ryu is
played by AI
that learned
by itself to
play
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Source: https://blogs.nvidia.com/blog/2016/07/29/whats-difference-
artificial-intelligence-machine-learning-deep-learning-ai/
The AI time line
A broad overview
3
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Source: https://blogs.nvidia.com/blog/2016/07/29/whats-difference-
artificial-intelligence-machine-learning-deep-learning-ai/
The AI time line
The starting point ...
a series of
if-then-
else rules
4
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Source:
https://wiki.bath.ac.uk/display/BISAI/Expert+System+in+a+Gaming+En
viroment
Expert system
Example: Arriving
decision to
choose a weapon
at any time by the
expert system
Rule behind
choosing mini gun
Rule behind
choosing missile
Rule behind
choosing defensive
pulse
5
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
6
Source: https://www.youtube.com/watch?v=eWSbIXSbMis
AI vs AI
The actions
by E Honda
were
determined
by rules
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Source: https://yachtharbour.com/news/venus-spotted-in-mallorca-
1887?src=news_view_page_bar
Expert system
Rules to determine ...
a yacht?
7
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Expert system
The limitations
•Expensive and time consuming!
Experts are never cheap
•Bad in handling sophisticated
sensory inputs (like signals,
images)
•Possible to make dumb decision
since it just goes through rules;
no common sense in the system
•System not easy to be updated
Source: https://blogs.nvidia.com/blog/2016/07/29/whats-difference-
artificial-intelligence-machine-learning-deep-learning-ai/
8
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Source: https://verhaert.com/difference-machine-learning-deep-
learning/
Features + classifier
A new solution to the rescue
•Feature: a number or a vector
that describes something about
the input
•Machine learning: classifier
learns the pattern between
features and output
9
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Source: https://towardsdatascience.com/face-recognition-how-
lbph-works-90ec258c3d6b
Features + classifier
Features?
•Face recognition
10
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Features + classifier
Issues?
•We need to design features
manually, through much trial and
error, with luck
•Simple features unlikely to work;
need complicated feature
extraction
•Classifiers used are generic (like
SVM)
11
Source: https://towardsdatascience.com/face-recognition-how-
lbph-works-90ec258c3d6b
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Source: https://www.guru99.com/machine-learning-vs-deep-
learning.html
Learning the features
Better performance?
•Instead of we deciding the features,
is it possible to get algorithm to
learn to extract the most
appropriate features by itself?
•Series of feature extractors?
•All the way from pixels to classifier,
learn by itself, layer by layer?
•Train all the layers together?
12
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Source: http://cs231n.stanford.edu
Deep in action
Part 1
conv ReLU conv ReLU pool conv ReLU
13
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Source: http://cs231n.stanford.edu
conv ReLU conv ReLU
pool conv ReLU pool
ship
car
airplane
truck
bird
FC
14
Deep in action
Part 2
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Source: https://commons.wikimedia.org/wiki/File:Typical_cnn.png
Learning the features
The idea behind convolutional
neural network (convnet)
Feature Extraction
Classification
15
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Source: http://www.image-
net.org/challenges/LSVRC/2012/results.html
From then on …
ILSVRC 2012
Rank Error - 5 Algorithm Team
1 0.153 Deep convolutional neural network University of Toronto
2 0.262 Features + Fisher vectors + linear classifiers ISI
3 0.270 Features + Fisher vectors + SVM Oxford VGG
4 0.271 Not specified XRCE/INRIA
5 0.300
Dense SIFT + colour SIFT + Fisher vectors +
SVM
University of Amsterdam
16
•ILSVRC: ImageNet Large Scale
Visual Recognition Challenge
•Challenge: Correctly classify
150,000 images, 1000
categories
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
The progress
So far
17
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
18
Source:
http://www.williammalone.com/articles/create-
html5-canvas-javascript-drawing-app/
Application
Three main categories
Source:
https://bitsandatoms.co/tag/reinforcement-
learning/
Source: http://blog.ss8.com
Create
Identify
Act
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
19
Volvo in-car cameras
Combat distraction and drunk
driving
•'Pro-active' driver monitoring
system
Source: https://www.caranddriver.com/news/a26893035/volvo-interior-cameras-
distraction-drunk-driving/
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
20
Aircraft
Monitoring ground handling
Source: https://assaia.com/tmc/
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
21
Aircraft
Monitoring check-in
Source: https://assaia.com/tmc/
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
22
Behaviour
identification
by Aipoly
Source: https://www.aipoly.com
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
23
Object detection
Auto referee?
•Track ball and players
•Report score based on the events
in scene
Source: https://arxiv.org/abs/2004.09927
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
24
To create
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
25
Drawing
A sketch
tree
river
rock
water
cloud
sky
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
26
Source: https://blogs.nvidia.com/blog/2019/03/18/gaugan-
photorealistic-landscapes-nvidia-research/
Drawing
A sketch
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
27
Source: https://blogs.nvidia.com/blog/2019/03/18/gaugan-
photorealistic-landscapes-nvidia-research/
Drawing
GauGAN (inspired by Gauguin)
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
28
Practical
usage
fashion
Source: https://qz.com/1090267/artificial-intelligence-can-now-
show-you-how-those-pants-will-fit/
A B C D
E F G H
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
29
Generative Adversarial
Networks
Change face
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
30
Generative Adversarial
Networks
Change face
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
31
Change face
Unlimited possibilities
Source: https://app.generative.photos
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
32
How about 3D?
Face?
Source: http://cvl-demos.cs.nott.ac.uk/vrn/index.php
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
33
To act
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
34
Source: https://towardsdatascience.com/using-deep-q-learning-in-fifa-
18-to-perfect-the-art-of-free-kicks-f2e4e979ee66
Freekick please
Fifa 18
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
35
Source: https://towardsdatascience.com/using-deep-q-learning-in-fifa-
18-to-perfect-the-art-of-free-kicks-f2e4e979ee66
Deep Reinforcement
learning
Fifa 18
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
36
Source: https://www.youtube.com/watch?v=FCivsotZEjk
Picking strawberries
no human
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
37
Source: https://www.tesla.com/autopilotAI
Autonomous
vehicle
Tesla autopilot
© 2021 National University of Singapore.
All Rights Reserved.
govtechbrbg01/v1.0
of 38
Conclusion •The AI that uses machine-learned
features (also known as deep
learning) is driving the current AI
boom
•Many applications not possible in
the past are made possible through
deep learning
•There are many uses of deep
learning to be imagined and
explored
38

More Related Content

Similar to The Rise of Machine-Learned Features

AI Blockchain IoT Convergence System Development
AI Blockchain IoT Convergence System DevelopmentAI Blockchain IoT Convergence System Development
AI Blockchain IoT Convergence System Development
Alex G. Lee, Ph.D. Esq. CLP
 

Similar to The Rise of Machine-Learned Features (20)

“How Transformers Are Changing the Nature of Deep Learning Models,” a Present...
“How Transformers Are Changing the Nature of Deep Learning Models,” a Present...“How Transformers Are Changing the Nature of Deep Learning Models,” a Present...
“How Transformers Are Changing the Nature of Deep Learning Models,” a Present...
 
Avinash thesis final
Avinash thesis finalAvinash thesis final
Avinash thesis final
 
TDD for jenkins pipelines
TDD for jenkins pipelinesTDD for jenkins pipelines
TDD for jenkins pipelines
 
Motion capture for Animation
Motion capture for AnimationMotion capture for Animation
Motion capture for Animation
 
Gluecon 2017 - GoMake | Flying Dreams: Real-Time Communication from the Edge ...
Gluecon 2017 - GoMake | Flying Dreams: Real-Time Communication from the Edge ...Gluecon 2017 - GoMake | Flying Dreams: Real-Time Communication from the Edge ...
Gluecon 2017 - GoMake | Flying Dreams: Real-Time Communication from the Edge ...
 
Genius Loci: iBeacons in the Academic Context
Genius Loci: iBeacons in the Academic ContextGenius Loci: iBeacons in the Academic Context
Genius Loci: iBeacons in the Academic Context
 
Year 5-6: Ideas for teaching coding
Year 5-6: Ideas for teaching codingYear 5-6: Ideas for teaching coding
Year 5-6: Ideas for teaching coding
 
Web Developers are now Mobile Developers
Web Developers are now Mobile Developers Web Developers are now Mobile Developers
Web Developers are now Mobile Developers
 
AI Blockchain IoT Convergence System Development
AI Blockchain IoT Convergence System DevelopmentAI Blockchain IoT Convergence System Development
AI Blockchain IoT Convergence System Development
 
Wireframe prototype - Mockups
Wireframe prototype - MockupsWireframe prototype - Mockups
Wireframe prototype - Mockups
 
VVC Project.pptx
VVC Project.pptxVVC Project.pptx
VVC Project.pptx
 
VVC Project.pptx
VVC Project.pptxVVC Project.pptx
VVC Project.pptx
 
Tolunay Mengi CV - English - 07.07.2022.pdf
Tolunay Mengi CV - English - 07.07.2022.pdfTolunay Mengi CV - English - 07.07.2022.pdf
Tolunay Mengi CV - English - 07.07.2022.pdf
 
Ee 323 project
Ee 323 projectEe 323 project
Ee 323 project
 
42
4242
42
 
Digital Fabrication Studio: 3D Scanning
Digital Fabrication Studio: 3D ScanningDigital Fabrication Studio: 3D Scanning
Digital Fabrication Studio: 3D Scanning
 
VVC_PPT.pptx
VVC_PPT.pptxVVC_PPT.pptx
VVC_PPT.pptx
 
Lesson learned from React native and Flutter
Lesson learned from React native and FlutterLesson learned from React native and Flutter
Lesson learned from React native and Flutter
 
When AOI meets AI
When AOI meets AIWhen AOI meets AI
When AOI meets AI
 
ESL report
ESL reportESL report
ESL report
 

More from NUS-ISS

More from NUS-ISS (20)

Designing Impactful Services and User Experience - Lim Wee Khee
Designing Impactful Services and User Experience - Lim Wee KheeDesigning Impactful Services and User Experience - Lim Wee Khee
Designing Impactful Services and User Experience - Lim Wee Khee
 
Upskilling the Evolving Workforce with Digital Fluency for Tomorrow's Challen...
Upskilling the Evolving Workforce with Digital Fluency for Tomorrow's Challen...Upskilling the Evolving Workforce with Digital Fluency for Tomorrow's Challen...
Upskilling the Evolving Workforce with Digital Fluency for Tomorrow's Challen...
 
How the World's Leading Independent Automotive Distributor is Reinventing Its...
How the World's Leading Independent Automotive Distributor is Reinventing Its...How the World's Leading Independent Automotive Distributor is Reinventing Its...
How the World's Leading Independent Automotive Distributor is Reinventing Its...
 
The Importance of Cybersecurity for Digital Transformation
The Importance of Cybersecurity for Digital TransformationThe Importance of Cybersecurity for Digital Transformation
The Importance of Cybersecurity for Digital Transformation
 
Architecting CX Measurement Frameworks and Ensuring CX Metrics are fit for Pu...
Architecting CX Measurement Frameworks and Ensuring CX Metrics are fit for Pu...Architecting CX Measurement Frameworks and Ensuring CX Metrics are fit for Pu...
Architecting CX Measurement Frameworks and Ensuring CX Metrics are fit for Pu...
 
Understanding GenAI/LLM and What is Google Offering - Felix Goh
Understanding GenAI/LLM and What is Google Offering - Felix GohUnderstanding GenAI/LLM and What is Google Offering - Felix Goh
Understanding GenAI/LLM and What is Google Offering - Felix Goh
 
Digital Product-Centric Enterprise and Enterprise Architecture - Tan Eng Tsze
Digital Product-Centric Enterprise and Enterprise Architecture - Tan Eng TszeDigital Product-Centric Enterprise and Enterprise Architecture - Tan Eng Tsze
Digital Product-Centric Enterprise and Enterprise Architecture - Tan Eng Tsze
 
Emerging & Future Technology - How to Prepare for the Next 10 Years of Radica...
Emerging & Future Technology - How to Prepare for the Next 10 Years of Radica...Emerging & Future Technology - How to Prepare for the Next 10 Years of Radica...
Emerging & Future Technology - How to Prepare for the Next 10 Years of Radica...
 
Beyond the Hype: What Generative AI Means for the Future of Work - Damien Cum...
Beyond the Hype: What Generative AI Means for the Future of Work - Damien Cum...Beyond the Hype: What Generative AI Means for the Future of Work - Damien Cum...
Beyond the Hype: What Generative AI Means for the Future of Work - Damien Cum...
 
Supply Chain Security for Containerised Workloads - Lee Chuk Munn
Supply Chain Security for Containerised Workloads - Lee Chuk MunnSupply Chain Security for Containerised Workloads - Lee Chuk Munn
Supply Chain Security for Containerised Workloads - Lee Chuk Munn
 
Future of Learning - Yap Aye Wee.pdf
Future of Learning - Yap Aye Wee.pdfFuture of Learning - Yap Aye Wee.pdf
Future of Learning - Yap Aye Wee.pdf
 
Future of Learning - Khoong Chan Meng
Future of Learning - Khoong Chan MengFuture of Learning - Khoong Chan Meng
Future of Learning - Khoong Chan Meng
 
Site Reliability Engineer (SRE), We Keep The Lights On 24/7
Site Reliability Engineer (SRE), We Keep The Lights On 24/7Site Reliability Engineer (SRE), We Keep The Lights On 24/7
Site Reliability Engineer (SRE), We Keep The Lights On 24/7
 
Product Management in The Trenches for a Cloud Service
Product Management in The Trenches for a Cloud ServiceProduct Management in The Trenches for a Cloud Service
Product Management in The Trenches for a Cloud Service
 
Overview of Data and Analytics Essentials and Foundations
Overview of Data and Analytics Essentials and FoundationsOverview of Data and Analytics Essentials and Foundations
Overview of Data and Analytics Essentials and Foundations
 
Predictive Analytics
Predictive AnalyticsPredictive Analytics
Predictive Analytics
 
Feature Engineering for IoT
Feature Engineering for IoTFeature Engineering for IoT
Feature Engineering for IoT
 
Master of Technology in Software Engineering
Master of Technology in Software EngineeringMaster of Technology in Software Engineering
Master of Technology in Software Engineering
 
Master of Technology in Enterprise Business Analytics
Master of Technology in Enterprise Business AnalyticsMaster of Technology in Enterprise Business Analytics
Master of Technology in Enterprise Business Analytics
 
Diagnosing Complex Problems Using System Archetypes
Diagnosing Complex Problems Using System ArchetypesDiagnosing Complex Problems Using System Archetypes
Diagnosing Complex Problems Using System Archetypes
 

Recently uploaded

Finding Java's Hidden Performance Traps @ DevoxxUK 2024
Finding Java's Hidden Performance Traps @ DevoxxUK 2024Finding Java's Hidden Performance Traps @ DevoxxUK 2024
Finding Java's Hidden Performance Traps @ DevoxxUK 2024
Victor Rentea
 
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
?#DUbAI#??##{{(☎️+971_581248768%)**%*]'#abortion pills for sale in dubai@
 

Recently uploaded (20)

presentation ICT roal in 21st century education
presentation ICT roal in 21st century educationpresentation ICT roal in 21st century education
presentation ICT roal in 21st century education
 
Apidays New York 2024 - Passkeys: Developing APIs to enable passwordless auth...
Apidays New York 2024 - Passkeys: Developing APIs to enable passwordless auth...Apidays New York 2024 - Passkeys: Developing APIs to enable passwordless auth...
Apidays New York 2024 - Passkeys: Developing APIs to enable passwordless auth...
 
MS Copilot expands with MS Graph connectors
MS Copilot expands with MS Graph connectorsMS Copilot expands with MS Graph connectors
MS Copilot expands with MS Graph connectors
 
FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024FWD Group - Insurer Innovation Award 2024
FWD Group - Insurer Innovation Award 2024
 
ICT role in 21st century education and its challenges
ICT role in 21st century education and its challengesICT role in 21st century education and its challenges
ICT role in 21st century education and its challenges
 
Mcleodganj Call Girls 🥰 8617370543 Service Offer VIP Hot Model
Mcleodganj Call Girls 🥰 8617370543 Service Offer VIP Hot ModelMcleodganj Call Girls 🥰 8617370543 Service Offer VIP Hot Model
Mcleodganj Call Girls 🥰 8617370543 Service Offer VIP Hot Model
 
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
Web Form Automation for Bonterra Impact Management (fka Social Solutions Apri...
 
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin WoodPolkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
 
Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...
 
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...
Biography Of Angeliki Cooney | Senior Vice President Life Sciences | Albany, ...
 
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ..."I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...
 
Finding Java's Hidden Performance Traps @ DevoxxUK 2024
Finding Java's Hidden Performance Traps @ DevoxxUK 2024Finding Java's Hidden Performance Traps @ DevoxxUK 2024
Finding Java's Hidden Performance Traps @ DevoxxUK 2024
 
Rising Above_ Dubai Floods and the Fortitude of Dubai International Airport.pdf
Rising Above_ Dubai Floods and the Fortitude of Dubai International Airport.pdfRising Above_ Dubai Floods and the Fortitude of Dubai International Airport.pdf
Rising Above_ Dubai Floods and the Fortitude of Dubai International Airport.pdf
 
CNIC Information System with Pakdata Cf In Pakistan
CNIC Information System with Pakdata Cf In PakistanCNIC Information System with Pakdata Cf In Pakistan
CNIC Information System with Pakdata Cf In Pakistan
 
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
 
Apidays New York 2024 - APIs in 2030: The Risk of Technological Sleepwalk by ...
Apidays New York 2024 - APIs in 2030: The Risk of Technological Sleepwalk by ...Apidays New York 2024 - APIs in 2030: The Risk of Technological Sleepwalk by ...
Apidays New York 2024 - APIs in 2030: The Risk of Technological Sleepwalk by ...
 
MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024
 
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
 
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, AdobeApidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
 
DEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
DEV meet-up UiPath Document Understanding May 7 2024 AmsterdamDEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
DEV meet-up UiPath Document Understanding May 7 2024 Amsterdam
 

The Rise of Machine-Learned Features

  • 1.
  • 2. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 2 Source: https://www.youtube.com/watch?v=eWSbIXSbMis AI vs AI Ryu is played by AI that learned by itself to play
  • 3. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Source: https://blogs.nvidia.com/blog/2016/07/29/whats-difference- artificial-intelligence-machine-learning-deep-learning-ai/ The AI time line A broad overview 3
  • 4. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Source: https://blogs.nvidia.com/blog/2016/07/29/whats-difference- artificial-intelligence-machine-learning-deep-learning-ai/ The AI time line The starting point ... a series of if-then- else rules 4
  • 5. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Source: https://wiki.bath.ac.uk/display/BISAI/Expert+System+in+a+Gaming+En viroment Expert system Example: Arriving decision to choose a weapon at any time by the expert system Rule behind choosing mini gun Rule behind choosing missile Rule behind choosing defensive pulse 5
  • 6. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 6 Source: https://www.youtube.com/watch?v=eWSbIXSbMis AI vs AI The actions by E Honda were determined by rules
  • 7. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Source: https://yachtharbour.com/news/venus-spotted-in-mallorca- 1887?src=news_view_page_bar Expert system Rules to determine ... a yacht? 7
  • 8. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Expert system The limitations •Expensive and time consuming! Experts are never cheap •Bad in handling sophisticated sensory inputs (like signals, images) •Possible to make dumb decision since it just goes through rules; no common sense in the system •System not easy to be updated Source: https://blogs.nvidia.com/blog/2016/07/29/whats-difference- artificial-intelligence-machine-learning-deep-learning-ai/ 8
  • 9. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Source: https://verhaert.com/difference-machine-learning-deep- learning/ Features + classifier A new solution to the rescue •Feature: a number or a vector that describes something about the input •Machine learning: classifier learns the pattern between features and output 9
  • 10. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Source: https://towardsdatascience.com/face-recognition-how- lbph-works-90ec258c3d6b Features + classifier Features? •Face recognition 10
  • 11. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Features + classifier Issues? •We need to design features manually, through much trial and error, with luck •Simple features unlikely to work; need complicated feature extraction •Classifiers used are generic (like SVM) 11 Source: https://towardsdatascience.com/face-recognition-how- lbph-works-90ec258c3d6b
  • 12. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Source: https://www.guru99.com/machine-learning-vs-deep- learning.html Learning the features Better performance? •Instead of we deciding the features, is it possible to get algorithm to learn to extract the most appropriate features by itself? •Series of feature extractors? •All the way from pixels to classifier, learn by itself, layer by layer? •Train all the layers together? 12
  • 13. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Source: http://cs231n.stanford.edu Deep in action Part 1 conv ReLU conv ReLU pool conv ReLU 13
  • 14. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Source: http://cs231n.stanford.edu conv ReLU conv ReLU pool conv ReLU pool ship car airplane truck bird FC 14 Deep in action Part 2
  • 15. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Source: https://commons.wikimedia.org/wiki/File:Typical_cnn.png Learning the features The idea behind convolutional neural network (convnet) Feature Extraction Classification 15
  • 16. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Source: http://www.image- net.org/challenges/LSVRC/2012/results.html From then on … ILSVRC 2012 Rank Error - 5 Algorithm Team 1 0.153 Deep convolutional neural network University of Toronto 2 0.262 Features + Fisher vectors + linear classifiers ISI 3 0.270 Features + Fisher vectors + SVM Oxford VGG 4 0.271 Not specified XRCE/INRIA 5 0.300 Dense SIFT + colour SIFT + Fisher vectors + SVM University of Amsterdam 16 •ILSVRC: ImageNet Large Scale Visual Recognition Challenge •Challenge: Correctly classify 150,000 images, 1000 categories
  • 17. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 The progress So far 17
  • 18. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 18 Source: http://www.williammalone.com/articles/create- html5-canvas-javascript-drawing-app/ Application Three main categories Source: https://bitsandatoms.co/tag/reinforcement- learning/ Source: http://blog.ss8.com Create Identify Act
  • 19. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 19 Volvo in-car cameras Combat distraction and drunk driving •'Pro-active' driver monitoring system Source: https://www.caranddriver.com/news/a26893035/volvo-interior-cameras- distraction-drunk-driving/
  • 20. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 20 Aircraft Monitoring ground handling Source: https://assaia.com/tmc/
  • 21. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 21 Aircraft Monitoring check-in Source: https://assaia.com/tmc/
  • 22. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 22 Behaviour identification by Aipoly Source: https://www.aipoly.com
  • 23. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 23 Object detection Auto referee? •Track ball and players •Report score based on the events in scene Source: https://arxiv.org/abs/2004.09927
  • 24. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 24 To create
  • 25. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 25 Drawing A sketch tree river rock water cloud sky
  • 26. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 26 Source: https://blogs.nvidia.com/blog/2019/03/18/gaugan- photorealistic-landscapes-nvidia-research/ Drawing A sketch
  • 27. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 27 Source: https://blogs.nvidia.com/blog/2019/03/18/gaugan- photorealistic-landscapes-nvidia-research/ Drawing GauGAN (inspired by Gauguin)
  • 28. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 28 Practical usage fashion Source: https://qz.com/1090267/artificial-intelligence-can-now- show-you-how-those-pants-will-fit/ A B C D E F G H
  • 29. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 29 Generative Adversarial Networks Change face
  • 30. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 30 Generative Adversarial Networks Change face
  • 31. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 31 Change face Unlimited possibilities Source: https://app.generative.photos
  • 32. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 32 How about 3D? Face? Source: http://cvl-demos.cs.nott.ac.uk/vrn/index.php
  • 33. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 33 To act
  • 34. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 34 Source: https://towardsdatascience.com/using-deep-q-learning-in-fifa- 18-to-perfect-the-art-of-free-kicks-f2e4e979ee66 Freekick please Fifa 18
  • 35. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 35 Source: https://towardsdatascience.com/using-deep-q-learning-in-fifa- 18-to-perfect-the-art-of-free-kicks-f2e4e979ee66 Deep Reinforcement learning Fifa 18
  • 36. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 36 Source: https://www.youtube.com/watch?v=FCivsotZEjk Picking strawberries no human
  • 37. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 37 Source: https://www.tesla.com/autopilotAI Autonomous vehicle Tesla autopilot
  • 38. © 2021 National University of Singapore. All Rights Reserved. govtechbrbg01/v1.0 of 38 Conclusion •The AI that uses machine-learned features (also known as deep learning) is driving the current AI boom •Many applications not possible in the past are made possible through deep learning •There are many uses of deep learning to be imagined and explored 38