SlideShare a Scribd company logo
1 of 30
1
Graph Neural Networks
2
Applications
3
Spotlight: A chemist’s tale
4
What’s this? Is a molecule a graph?
If so, what are its nodes,
edges, features?
Predict if a molecule is a potent drug
5
1. Train a GNN on a curated dataset where response is known
2. Once trained, use the model and apply any molecule
3. Select ~top-100 candidates
4. Get chemists to thoroughly investigate them
Halicin - a powerful drug discovered by GNN in
2020
“We wanted to develop a platform that would allow us to harness the power of artificial intelligence to usher in a new age of antibiotic
drug discovery. Our approach revealed this amazing molecule which is arguably one of the more powerful antibiotics that has been
discovered.” - James Collins, Professor of Medical Engineering and Science, MIT
6
6
Success stories of GNNs
7
Machine Learning problems using Graph data
8
?
Node-level predictions
Does this person smoke?
(Unlabelled node)
Link-level predictions
Next Prime video?
Graph-level predictions
Is this molecule a suitable
drug?
Concept of Embeddings
9
Embeddings
10
banana
mango
dog
Vocabulary size (V)
banana
mango
dog
D
Local representation (1-hot) Distributed representation
Graph Neural
Networks -
Intuition (and
some Math)
11
A simplified view
12
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
GNN
h1 h2 h3 …
h1 h2 h3 …
h1 h2 h3 …
h1 h2 h3 …
h1 h2 h3 …
Vectorizing this…
13
Inputs(X, A) Latents(H, A)
GNN
Message passing
14
1
2
3
4
5
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
AGGREGATE UPDATE
1
2
3
4
5
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
More Message passing
15
1
2
3
4
5
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
1
2
3
4
5
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
[...]
Multiple hops of message passing
16
1
2
3
4
5
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
1
2
3
4
5
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
x1 x2
x
3
…
What to do with GNN outputs?
17
Inputs(X, A) Latents(H, A)
GNN
Graph Neural Networks - more Math
18
Two-step message passing: a recap
19
1
2
3
4
5
Target
node
1
2
3
4
AGGREGATE
1. AGGREGATE - pass information (the “message”) from the target node’s neighbours to
the target node
2. UPDATE - update each node’s features based on “message” to form an embedded
representation
Generic form of message passing
20
1
2
3
4
AGGREGATE
1
2
3
4
5
Target node
h = node features or embeddings
k = number of hops
Using neural networks for aggregate and update
21
Each node’s updated value becomes a weighting of its previous value + a weighting of its
neighbours’ values
Make message passing more efficient by
simplifying, generalizing and sharing parameters
22
Collapse the two weight vectors into W by adding self-loops to the adjacency matrix
Base GNNs on a “convolution” perspective
23
Normalize by # of nodes in the
neighbourhood
“Original” GNN (2009)
GCN (2016)
Graph Neural
Networks -
Implementation
notes
24
25
pyTorch and pyG
1.
A wealth of libraries
You can mix-and-match some of these libraries to train and predict node/edge/graph
classification problems.
tensorflow
2.
GDS from neo4j
3.
25
Book recommendation
26
Thankyou!
Kumar Sankara Iyer
Engineering Director, Thoughtworks
www.linkedin.com/in/aquaraga
27
Appendix
28
Neural Networks - you’ve seen this diagram a
LOT!
29
Neural networks: weights, biases, activations
30

More Related Content

What's hot

Faster R-CNN: Towards real-time object detection with region proposal network...
Faster R-CNN: Towards real-time object detection with region proposal network...Faster R-CNN: Towards real-time object detection with region proposal network...
Faster R-CNN: Towards real-time object detection with region proposal network...Universitat Politècnica de Catalunya
 
Introduction to Graph neural networks @ Vienna Deep Learning meetup
Introduction to Graph neural networks @  Vienna Deep Learning meetupIntroduction to Graph neural networks @  Vienna Deep Learning meetup
Introduction to Graph neural networks @ Vienna Deep Learning meetupLiad Magen
 
Transformers in Vision: From Zero to Hero
Transformers in Vision: From Zero to HeroTransformers in Vision: From Zero to Hero
Transformers in Vision: From Zero to HeroBill Liu
 
Handwritten Digit Recognition using Convolutional Neural Networks
Handwritten Digit Recognition using Convolutional Neural  NetworksHandwritten Digit Recognition using Convolutional Neural  Networks
Handwritten Digit Recognition using Convolutional Neural NetworksIRJET Journal
 
Lec14: Evaluation Framework for Medical Image Segmentation
Lec14: Evaluation Framework for Medical Image SegmentationLec14: Evaluation Framework for Medical Image Segmentation
Lec14: Evaluation Framework for Medical Image SegmentationUlaş Bağcı
 
Convolutional Neural Network - CNN | How CNN Works | Deep Learning Course | S...
Convolutional Neural Network - CNN | How CNN Works | Deep Learning Course | S...Convolutional Neural Network - CNN | How CNN Works | Deep Learning Course | S...
Convolutional Neural Network - CNN | How CNN Works | Deep Learning Course | S...Simplilearn
 
Image Segmentation (D3L1 2017 UPC Deep Learning for Computer Vision)
Image Segmentation (D3L1 2017 UPC Deep Learning for Computer Vision)Image Segmentation (D3L1 2017 UPC Deep Learning for Computer Vision)
Image Segmentation (D3L1 2017 UPC Deep Learning for Computer Vision)Universitat Politècnica de Catalunya
 
Activation functions and Training Algorithms for Deep Neural network
Activation functions and Training Algorithms for Deep Neural networkActivation functions and Training Algorithms for Deep Neural network
Activation functions and Training Algorithms for Deep Neural networkGayatri Khanvilkar
 
210523 swin transformer v1.5
210523 swin transformer v1.5210523 swin transformer v1.5
210523 swin transformer v1.5taeseon ryu
 
Deep learning for medical imaging
Deep learning for medical imagingDeep learning for medical imaging
Deep learning for medical imaginggeetachauhan
 
Image-to-Image Translation with Conditional Adversarial Nets (UPC Reading Group)
Image-to-Image Translation with Conditional Adversarial Nets (UPC Reading Group)Image-to-Image Translation with Conditional Adversarial Nets (UPC Reading Group)
Image-to-Image Translation with Conditional Adversarial Nets (UPC Reading Group)Universitat Politècnica de Catalunya
 
Image classification using convolutional neural network
Image classification using convolutional neural networkImage classification using convolutional neural network
Image classification using convolutional neural networkKIRAN R
 
Image classification with Deep Neural Networks
Image classification with Deep Neural NetworksImage classification with Deep Neural Networks
Image classification with Deep Neural NetworksYogendra Tamang
 
Classification Based Machine Learning Algorithms
Classification Based Machine Learning AlgorithmsClassification Based Machine Learning Algorithms
Classification Based Machine Learning AlgorithmsMd. Main Uddin Rony
 
Convolutional Neural Network
Convolutional Neural NetworkConvolutional Neural Network
Convolutional Neural NetworkVignesh Suresh
 
Image classification using cnn
Image classification using cnnImage classification using cnn
Image classification using cnnSumeraHangi
 
Object Detection using Deep Neural Networks
Object Detection using Deep Neural NetworksObject Detection using Deep Neural Networks
Object Detection using Deep Neural NetworksUsman Qayyum
 

What's hot (20)

Faster R-CNN: Towards real-time object detection with region proposal network...
Faster R-CNN: Towards real-time object detection with region proposal network...Faster R-CNN: Towards real-time object detection with region proposal network...
Faster R-CNN: Towards real-time object detection with region proposal network...
 
Introduction to Graph neural networks @ Vienna Deep Learning meetup
Introduction to Graph neural networks @  Vienna Deep Learning meetupIntroduction to Graph neural networks @  Vienna Deep Learning meetup
Introduction to Graph neural networks @ Vienna Deep Learning meetup
 
Zero shot learning
Zero shot learning Zero shot learning
Zero shot learning
 
Transformers in Vision: From Zero to Hero
Transformers in Vision: From Zero to HeroTransformers in Vision: From Zero to Hero
Transformers in Vision: From Zero to Hero
 
Handwritten Digit Recognition using Convolutional Neural Networks
Handwritten Digit Recognition using Convolutional Neural  NetworksHandwritten Digit Recognition using Convolutional Neural  Networks
Handwritten Digit Recognition using Convolutional Neural Networks
 
Lec14: Evaluation Framework for Medical Image Segmentation
Lec14: Evaluation Framework for Medical Image SegmentationLec14: Evaluation Framework for Medical Image Segmentation
Lec14: Evaluation Framework for Medical Image Segmentation
 
Convolutional Neural Network - CNN | How CNN Works | Deep Learning Course | S...
Convolutional Neural Network - CNN | How CNN Works | Deep Learning Course | S...Convolutional Neural Network - CNN | How CNN Works | Deep Learning Course | S...
Convolutional Neural Network - CNN | How CNN Works | Deep Learning Course | S...
 
GoogLeNet Insights
GoogLeNet InsightsGoogLeNet Insights
GoogLeNet Insights
 
Image Segmentation (D3L1 2017 UPC Deep Learning for Computer Vision)
Image Segmentation (D3L1 2017 UPC Deep Learning for Computer Vision)Image Segmentation (D3L1 2017 UPC Deep Learning for Computer Vision)
Image Segmentation (D3L1 2017 UPC Deep Learning for Computer Vision)
 
Activation functions and Training Algorithms for Deep Neural network
Activation functions and Training Algorithms for Deep Neural networkActivation functions and Training Algorithms for Deep Neural network
Activation functions and Training Algorithms for Deep Neural network
 
210523 swin transformer v1.5
210523 swin transformer v1.5210523 swin transformer v1.5
210523 swin transformer v1.5
 
Deep learning for medical imaging
Deep learning for medical imagingDeep learning for medical imaging
Deep learning for medical imaging
 
cnn ppt.pptx
cnn ppt.pptxcnn ppt.pptx
cnn ppt.pptx
 
Image-to-Image Translation with Conditional Adversarial Nets (UPC Reading Group)
Image-to-Image Translation with Conditional Adversarial Nets (UPC Reading Group)Image-to-Image Translation with Conditional Adversarial Nets (UPC Reading Group)
Image-to-Image Translation with Conditional Adversarial Nets (UPC Reading Group)
 
Image classification using convolutional neural network
Image classification using convolutional neural networkImage classification using convolutional neural network
Image classification using convolutional neural network
 
Image classification with Deep Neural Networks
Image classification with Deep Neural NetworksImage classification with Deep Neural Networks
Image classification with Deep Neural Networks
 
Classification Based Machine Learning Algorithms
Classification Based Machine Learning AlgorithmsClassification Based Machine Learning Algorithms
Classification Based Machine Learning Algorithms
 
Convolutional Neural Network
Convolutional Neural NetworkConvolutional Neural Network
Convolutional Neural Network
 
Image classification using cnn
Image classification using cnnImage classification using cnn
Image classification using cnn
 
Object Detection using Deep Neural Networks
Object Detection using Deep Neural NetworksObject Detection using Deep Neural Networks
Object Detection using Deep Neural Networks
 

Similar to Graph Neural Networks.pptx

TensorFlow London: Cutting edge generative models
TensorFlow London: Cutting edge generative modelsTensorFlow London: Cutting edge generative models
TensorFlow London: Cutting edge generative modelsSeldon
 
CEHS 2016 Poster
CEHS 2016 PosterCEHS 2016 Poster
CEHS 2016 PosterEric Ma
 
A Fuzzy Approach For Multi-Domain Sentiment Analysis
A Fuzzy Approach For Multi-Domain Sentiment AnalysisA Fuzzy Approach For Multi-Domain Sentiment Analysis
A Fuzzy Approach For Multi-Domain Sentiment AnalysisMauro Dragoni
 
2017 07 03_meetup_d
2017 07 03_meetup_d2017 07 03_meetup_d
2017 07 03_meetup_dDana Brophy
 
2017 07 03_meetup_d
2017 07 03_meetup_d2017 07 03_meetup_d
2017 07 03_meetup_dDana Brophy
 
Large Scale Data Clustering: an overview
Large Scale Data Clustering: an overviewLarge Scale Data Clustering: an overview
Large Scale Data Clustering: an overviewVahid Mirjalili
 
Diagnosis Chest Diseases Using Neural Network and Genetic Hybrid Algorithm
Diagnosis Chest Diseases Using Neural Network and Genetic Hybrid AlgorithmDiagnosis Chest Diseases Using Neural Network and Genetic Hybrid Algorithm
Diagnosis Chest Diseases Using Neural Network and Genetic Hybrid AlgorithmIJERA Editor
 
Towards Statistical Queries over Distributed Private User Data
Towards Statistical Queries over Distributed Private User Data Towards Statistical Queries over Distributed Private User Data
Towards Statistical Queries over Distributed Private User Data Serafeim Chatzopoulos
 
A Survey of Deep Learning Algorithms for Malware Detection
A Survey of Deep Learning Algorithms for Malware DetectionA Survey of Deep Learning Algorithms for Malware Detection
A Survey of Deep Learning Algorithms for Malware DetectionIJCSIS Research Publications
 
Deep learning in oil and gas
Deep learning in oil and gasDeep learning in oil and gas
Deep learning in oil and gasYang Cong
 

Similar to Graph Neural Networks.pptx (20)

TensorFlow London: Cutting edge generative models
TensorFlow London: Cutting edge generative modelsTensorFlow London: Cutting edge generative models
TensorFlow London: Cutting edge generative models
 
Introduction to Deep learning
Introduction to Deep learningIntroduction to Deep learning
Introduction to Deep learning
 
CEHS 2016 Poster
CEHS 2016 PosterCEHS 2016 Poster
CEHS 2016 Poster
 
A Fuzzy Approach For Multi-Domain Sentiment Analysis
A Fuzzy Approach For Multi-Domain Sentiment AnalysisA Fuzzy Approach For Multi-Domain Sentiment Analysis
A Fuzzy Approach For Multi-Domain Sentiment Analysis
 
2017 07 03_meetup_d
2017 07 03_meetup_d2017 07 03_meetup_d
2017 07 03_meetup_d
 
2017 07 03_meetup_d
2017 07 03_meetup_d2017 07 03_meetup_d
2017 07 03_meetup_d
 
Building the next generation of statistical tools for outbreak response using R
Building the next generation of statistical tools for outbreak response using RBuilding the next generation of statistical tools for outbreak response using R
Building the next generation of statistical tools for outbreak response using R
 
ProjectReport
ProjectReportProjectReport
ProjectReport
 
XGBoostLSS - An extension of XGBoost to probabilistic forecasting, Alexander ...
XGBoostLSS - An extension of XGBoost to probabilistic forecasting, Alexander ...XGBoostLSS - An extension of XGBoost to probabilistic forecasting, Alexander ...
XGBoostLSS - An extension of XGBoost to probabilistic forecasting, Alexander ...
 
323462348
323462348323462348
323462348
 
323462348
323462348323462348
323462348
 
Large Scale Data Clustering: an overview
Large Scale Data Clustering: an overviewLarge Scale Data Clustering: an overview
Large Scale Data Clustering: an overview
 
Esa act-2020 la mura
Esa act-2020 la muraEsa act-2020 la mura
Esa act-2020 la mura
 
Diagnosis Chest Diseases Using Neural Network and Genetic Hybrid Algorithm
Diagnosis Chest Diseases Using Neural Network and Genetic Hybrid AlgorithmDiagnosis Chest Diseases Using Neural Network and Genetic Hybrid Algorithm
Diagnosis Chest Diseases Using Neural Network and Genetic Hybrid Algorithm
 
Towards Statistical Queries over Distributed Private User Data
Towards Statistical Queries over Distributed Private User Data Towards Statistical Queries over Distributed Private User Data
Towards Statistical Queries over Distributed Private User Data
 
A Survey of Deep Learning Algorithms for Malware Detection
A Survey of Deep Learning Algorithms for Malware DetectionA Survey of Deep Learning Algorithms for Malware Detection
A Survey of Deep Learning Algorithms for Malware Detection
 
Major project.pptx
Major project.pptxMajor project.pptx
Major project.pptx
 
paper_148.pptx
paper_148.pptxpaper_148.pptx
paper_148.pptx
 
20181212 ibm aot
20181212 ibm aot20181212 ibm aot
20181212 ibm aot
 
Deep learning in oil and gas
Deep learning in oil and gasDeep learning in oil and gas
Deep learning in oil and gas
 

Recently uploaded

Bun (KitWorks Team Study 노별마루 발표 2024.4.22)
Bun (KitWorks Team Study 노별마루 발표 2024.4.22)Bun (KitWorks Team Study 노별마루 발표 2024.4.22)
Bun (KitWorks Team Study 노별마루 발표 2024.4.22)Wonjun Hwang
 
AI as an Interface for Commercial Buildings
AI as an Interface for Commercial BuildingsAI as an Interface for Commercial Buildings
AI as an Interface for Commercial BuildingsMemoori
 
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024BookNet Canada
 
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmaticsKotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmaticscarlostorres15106
 
Build your next Gen AI Breakthrough - April 2024
Build your next Gen AI Breakthrough - April 2024Build your next Gen AI Breakthrough - April 2024
Build your next Gen AI Breakthrough - April 2024Neo4j
 
New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024BookNet Canada
 
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024BookNet Canada
 
SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024Scott Keck-Warren
 
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationBeyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationSafe Software
 
Designing IA for AI - Information Architecture Conference 2024
Designing IA for AI - Information Architecture Conference 2024Designing IA for AI - Information Architecture Conference 2024
Designing IA for AI - Information Architecture Conference 2024Enterprise Knowledge
 
My Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 PresentationMy Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 PresentationRidwan Fadjar
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 3652toLead Limited
 
Are Multi-Cloud and Serverless Good or Bad?
Are Multi-Cloud and Serverless Good or Bad?Are Multi-Cloud and Serverless Good or Bad?
Are Multi-Cloud and Serverless Good or Bad?Mattias Andersson
 
Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024BookNet Canada
 
APIForce Zurich 5 April Automation LPDG
APIForce Zurich 5 April  Automation LPDGAPIForce Zurich 5 April  Automation LPDG
APIForce Zurich 5 April Automation LPDGMarianaLemus7
 
Swan(sea) Song – personal research during my six years at Swansea ... and bey...
Swan(sea) Song – personal research during my six years at Swansea ... and bey...Swan(sea) Song – personal research during my six years at Swansea ... and bey...
Swan(sea) Song – personal research during my six years at Swansea ... and bey...Alan Dix
 
costume and set research powerpoint presentation
costume and set research powerpoint presentationcostume and set research powerpoint presentation
costume and set research powerpoint presentationphoebematthew05
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesSinan KOZAK
 
Streamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupStreamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupFlorian Wilhelm
 

Recently uploaded (20)

Bun (KitWorks Team Study 노별마루 발표 2024.4.22)
Bun (KitWorks Team Study 노별마루 발표 2024.4.22)Bun (KitWorks Team Study 노별마루 발표 2024.4.22)
Bun (KitWorks Team Study 노별마루 발표 2024.4.22)
 
AI as an Interface for Commercial Buildings
AI as an Interface for Commercial BuildingsAI as an Interface for Commercial Buildings
AI as an Interface for Commercial Buildings
 
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
 
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmaticsKotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
 
Build your next Gen AI Breakthrough - April 2024
Build your next Gen AI Breakthrough - April 2024Build your next Gen AI Breakthrough - April 2024
Build your next Gen AI Breakthrough - April 2024
 
New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
 
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
 
SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024
 
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationBeyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
 
Designing IA for AI - Information Architecture Conference 2024
Designing IA for AI - Information Architecture Conference 2024Designing IA for AI - Information Architecture Conference 2024
Designing IA for AI - Information Architecture Conference 2024
 
My Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 PresentationMy Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 Presentation
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
 
Are Multi-Cloud and Serverless Good or Bad?
Are Multi-Cloud and Serverless Good or Bad?Are Multi-Cloud and Serverless Good or Bad?
Are Multi-Cloud and Serverless Good or Bad?
 
Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
 
APIForce Zurich 5 April Automation LPDG
APIForce Zurich 5 April  Automation LPDGAPIForce Zurich 5 April  Automation LPDG
APIForce Zurich 5 April Automation LPDG
 
Swan(sea) Song – personal research during my six years at Swansea ... and bey...
Swan(sea) Song – personal research during my six years at Swansea ... and bey...Swan(sea) Song – personal research during my six years at Swansea ... and bey...
Swan(sea) Song – personal research during my six years at Swansea ... and bey...
 
Vulnerability_Management_GRC_by Sohang Sengupta.pptx
Vulnerability_Management_GRC_by Sohang Sengupta.pptxVulnerability_Management_GRC_by Sohang Sengupta.pptx
Vulnerability_Management_GRC_by Sohang Sengupta.pptx
 
costume and set research powerpoint presentation
costume and set research powerpoint presentationcostume and set research powerpoint presentation
costume and set research powerpoint presentation
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen Frames
 
Streamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupStreamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project Setup
 

Graph Neural Networks.pptx

Editor's Notes

  1. Set expectation for the session You should be able to understand why people use GNNs You should have a 20000-foot idea of what they actually do - at least develop an intuition of what they do Appreciate the math (if you have some exposure to Neural networks) Overall we will be setting the stage for a subsequent workshop (perhaps in the next neo4j meetup) where we will do some hands-on. Must-watch: https://www.youtube.com/watch?v=cWIeTMklzNg
  2. Just like graph databases are found everywhere, predictions based on graphs also have their applications. From a social graph of LinkedIn, you can suggest new connections. Could be used in e-commerce recommender systems Could be used in medicine/pharmacy - say is this drug potent? Could be used in social networks - say providing connection recommendations https://www.youtube.com/watch?v=fOctJB4kVlM&list=PLV8yxwGOxvvoNkzPfCx2i8an--Tkt7O8Z&index=1 2:14
  3. Molecule Nodes - atoms Edges - bonds Features of nodes - atom types, number of protons, charge Features of edges - bond type
  4. One you have a molecule represented as a graph, you can train a Graph Neural Network to perform - say a binary task. Will this molecule inhibit a given bacteria? First, you will have a curated dataset of several molecules and you know the outcome - i.e. whether it will inhibit E-coli or not. You train a Graph neural network with this training data. Once you have trained the GNN, you can feed it a bunch of candidate molecules. Take the most promising N number of molecules.
  5. https://news.mit.edu/2020/artificial-intelligence-identifies-new-antibiotic-0220
  6. Halicin in 2020 GNNs in 1990s in the chemical industry - a bit of history
  7. So what can you do with graph data? Node-level predictions - If you have a graph with unlabelled nodes, you want to predict attributes about them or you want to classify them. GNNs as you’ll see soon will use information from other nodes to infer on these unlabelled nodes Edge-prediction - Typically used by companies to predict which product will be purchased next by a customer. Here, it can used to predict if there will be a link between a person node and an item node Graph-level prediction - If a molecule is a potent drug or not https://www.youtube.com/watch?v=fOctJB4kVlM&list=PLV8yxwGOxvvoNkzPfCx2i8an--Tkt7O8Z&index=2 3:20
  8. This is a commonly encountered concept in NLP and deep learning. How do you represent discrete things in a large set, say the words in a dictionary? We have a huge sparse vector that has mostly 0s. It just has 1 for the word under consideration at the appropriate index. Problem is that we don’t know how they will correlate. Most deep learning models tend to learn a distributed vector representation. In this case, you can see that “banana” and “mango” have some similarity - maybe because they are both fruits and are yellow. So V could be 10000, and D maybe some 128 or 500 or something like that. In a way, it compresses as well. Distributed Vector representations - https://www.youtube.com/watch?v=zCEYiCxrL_0 5:29
  9. You start with a graph - modelling decision. It should encode the problem. For each of the nodes, there could be some information - either an embedding or a set of features. The output of a GNN is the same graph. For each node, there is a vector representation. Each node now should have the information not simply about itself - but rather how it belongs in the context of the graph. Training and recommendation
  10. Recommendation
  11. The aggregate function should be permutation invariant. Training
  12. Training Too much message passing - over-smoothing
  13. In the first step, the nodes nly know about themselves In the next step, all of them know about their neighbours. Then their neighbours’ neighbours. In each step, the perceptive field increases. Finally, every node would have learnt how it “belongs” in the graph.
  14. Once you have these vectors, you can do: Node classification - classify each node independently by applying a shared layer Link classification Aggregate all the h vectors 12:59 https://www.youtube.com/watch?v=8owQBFAHw7E Recommendation
  15. Show n layers of convolution
  16. Aggregate function needs to be permutation invariant. For eg; if there is a new neighbour, the output shape (whether it is a number or a matrix) should not change.
  17. Summation is a choice - you could either do mean, max, min, etc.
  18. Do not treat target and neighbour nodes differently. Adjacency matrix defines
  19. pyTorch - From Meta, now under Linux foundation. Optimized tensor library for deep learning using GPUs and CPUs. pyG - Has extension for graph learning. Supports GNN architectures like GCN and GAT Tensorflow GDS - Random walks with restarts sampling (server-side and client-side components, uses Apache arrow for in-memory analytics)