SlideShare a Scribd company logo
Theory and Evaluation Metrics for
Learning Disentangled Representations
Kien Do and Truyen Tran
Applied AI Institute (A2I2), Deakin University, Australia
12020
Examples about disentangled representations
2
Overcoming the Disentanglement vs Reconstruction Trade-off via Jacobian Supervision (Lezama et. al., ICLR-2019)
Current challenges
• The mathematic interpretation of disentangled representations is still
not clear
• Most evaluations metrics are heuristic and not robust
3
Our goals
• Provide a mathematically formal definition of disentangled
representations
• Propose proper, theoretically driven metrics for fair evaluations
4
An informal definition
Disentanglement: A process of decorrelating data into separate pieces
of information, each of which corresponds to a predefined concept.
This suggests 3 main properties of disentangled representations:
• Informativeness: Large enough pieces of information
• Separability: Pieces of information are not overlapped
• Interpretability: Each piece of information corresponds to a concept
6
Mathematic formulas of these properties
• The informativeness of is defined by:
• are separable w.r.t. if:
• are fully separable (independent) if:
• is fully interpretable w.r.t. if:
7
Definition of disentangled representations
8
Compared to other related works
• Higgins et. al., 2018 [1] attempted to define disentanglement based
on group theory. Their definition shares similar observations as ours
but is less convenient for probabilistic model design.
• Eastwood et. al., 2018 [2] did NOT provide explicit definition of
disentangled representations but characterized them along 3
dimensions: disentanglement, compactness, and informativeness.
• The main drawback of disentanglement and compactness is that they require
knowledge of ALL ground truth factors of variation.
9
[1] Towards a definition of disentangled representations (Higgins et. al., ICLR-2018)
[2] A framework for the quantitative evaluation of disentangled representations (Eastwood et. al., ICLR-2018)
Our proposed metrics
• Informativeness:
• Separability + Informativeness (no labeled factors available):
10
where
Our proposed metrics (cont.)
• Separability + Interpretability (labeled factors provided):
where and is the highest and second highest MI
values between every and .
11
Advantages of our metrics
• Support both supervised/unsupervised models
• Can be applied for real datasets
• Do not require any training procedure
• Provide consistent results across different models
• Agree with visual results
12
Experiments and Results
Please refer to our paper!
13
14
Thank you for listening!

More Related Content

What's hot

WXGB6108_Article Review_The Effect of Attitudes, Goal Setting and Self-Effica...
WXGB6108_Article Review_The Effect of Attitudes, Goal Setting and Self-Effica...WXGB6108_Article Review_The Effect of Attitudes, Goal Setting and Self-Effica...
WXGB6108_Article Review_The Effect of Attitudes, Goal Setting and Self-Effica...
Husna Zayadi
 
T OWARDS A S YSTEM D YNAMICS M ODELING M E- THOD B ASED ON DEMATEL
T OWARDS A  S YSTEM  D YNAMICS  M ODELING  M E- THOD B ASED ON  DEMATELT OWARDS A  S YSTEM  D YNAMICS  M ODELING  M E- THOD B ASED ON  DEMATEL
T OWARDS A S YSTEM D YNAMICS M ODELING M E- THOD B ASED ON DEMATEL
ijcsit
 
Machine Learning part 2 - Introduction to Data Science
Machine Learning part 2 -  Introduction to Data Science Machine Learning part 2 -  Introduction to Data Science
Machine Learning part 2 - Introduction to Data Science
Frank Kienle
 
Comparison between ER Modeling and Dimension Modeling
Comparison between ER Modeling and Dimension ModelingComparison between ER Modeling and Dimension Modeling
Comparison between ER Modeling and Dimension Modeling
Karuna Kak
 
Building classification model, tree model, confusion matrix and prediction ac...
Building classification model, tree model, confusion matrix and prediction ac...Building classification model, tree model, confusion matrix and prediction ac...
Building classification model, tree model, confusion matrix and prediction ac...
National Cheng Kung University
 
Student Performance Data Mining Project Report
Student Performance Data Mining Project ReportStudent Performance Data Mining Project Report
Student Performance Data Mining Project Report
Jinnah University for Women
 
Design pattern
Design patternDesign pattern
Design pattern
Jasdhir Singh
 
Decision tree
Decision treeDecision tree
Decision tree
Estiak Khan
 
Machine Learning part 3 - Introduction to data science
Machine Learning part 3 - Introduction to data science Machine Learning part 3 - Introduction to data science
Machine Learning part 3 - Introduction to data science
Frank Kienle
 
Classification of data using semi supervised learning a learning disability ...
Classification of data using semi supervised learning  a learning disability ...Classification of data using semi supervised learning  a learning disability ...
Classification of data using semi supervised learning a learning disability ...
IAEME Publication
 
A New Active Learning Technique Using Furthest Nearest Neighbour Criterion fo...
A New Active Learning Technique Using Furthest Nearest Neighbour Criterion fo...A New Active Learning Technique Using Furthest Nearest Neighbour Criterion fo...
A New Active Learning Technique Using Furthest Nearest Neighbour Criterion fo...
ijcsa
 
Been Kim - Interpretable machine learning, Nov 2015
Been Kim - Interpretable machine learning, Nov 2015Been Kim - Interpretable machine learning, Nov 2015
Been Kim - Interpretable machine learning, Nov 2015
Seattle DAML meetup
 
IRJET- Facial Emotion Detection using Convolutional Neural Network
IRJET- Facial Emotion Detection using Convolutional Neural NetworkIRJET- Facial Emotion Detection using Convolutional Neural Network
IRJET- Facial Emotion Detection using Convolutional Neural Network
IRJET Journal
 
Data Clustering in Education for Students
Data Clustering in Education for StudentsData Clustering in Education for Students
Data Clustering in Education for Students
IRJET Journal
 
Predicting the Presence of Learning Motivation in Electronic Learning: A New ...
Predicting the Presence of Learning Motivation in Electronic Learning: A New ...Predicting the Presence of Learning Motivation in Electronic Learning: A New ...
Predicting the Presence of Learning Motivation in Electronic Learning: A New ...
TELKOMNIKA JOURNAL
 
Explainable AI in Healthcare
Explainable AI in HealthcareExplainable AI in Healthcare
Explainable AI in Healthcare
vonaurum
 
TS4-3: Takumi Sato from Nagoya Institute of Technology
TS4-3: Takumi Sato from Nagoya Institute of TechnologyTS4-3: Takumi Sato from Nagoya Institute of Technology
TS4-3: Takumi Sato from Nagoya Institute of Technology
Jawad Haqbeen
 
Hybrid Classifier for Sentiment Analysis using Effective Pipelining
Hybrid Classifier for Sentiment Analysis using Effective PipeliningHybrid Classifier for Sentiment Analysis using Effective Pipelining
Hybrid Classifier for Sentiment Analysis using Effective Pipelining
IRJET Journal
 
Introduction to Interpretable Machine Learning
Introduction to Interpretable Machine LearningIntroduction to Interpretable Machine Learning
Introduction to Interpretable Machine Learning
Nguyen Giang
 
Model Evaluation in the land of Deep Learning
Model Evaluation in the land of Deep LearningModel Evaluation in the land of Deep Learning
Model Evaluation in the land of Deep Learning
Pramit Choudhary
 

What's hot (20)

WXGB6108_Article Review_The Effect of Attitudes, Goal Setting and Self-Effica...
WXGB6108_Article Review_The Effect of Attitudes, Goal Setting and Self-Effica...WXGB6108_Article Review_The Effect of Attitudes, Goal Setting and Self-Effica...
WXGB6108_Article Review_The Effect of Attitudes, Goal Setting and Self-Effica...
 
T OWARDS A S YSTEM D YNAMICS M ODELING M E- THOD B ASED ON DEMATEL
T OWARDS A  S YSTEM  D YNAMICS  M ODELING  M E- THOD B ASED ON  DEMATELT OWARDS A  S YSTEM  D YNAMICS  M ODELING  M E- THOD B ASED ON  DEMATEL
T OWARDS A S YSTEM D YNAMICS M ODELING M E- THOD B ASED ON DEMATEL
 
Machine Learning part 2 - Introduction to Data Science
Machine Learning part 2 -  Introduction to Data Science Machine Learning part 2 -  Introduction to Data Science
Machine Learning part 2 - Introduction to Data Science
 
Comparison between ER Modeling and Dimension Modeling
Comparison between ER Modeling and Dimension ModelingComparison between ER Modeling and Dimension Modeling
Comparison between ER Modeling and Dimension Modeling
 
Building classification model, tree model, confusion matrix and prediction ac...
Building classification model, tree model, confusion matrix and prediction ac...Building classification model, tree model, confusion matrix and prediction ac...
Building classification model, tree model, confusion matrix and prediction ac...
 
Student Performance Data Mining Project Report
Student Performance Data Mining Project ReportStudent Performance Data Mining Project Report
Student Performance Data Mining Project Report
 
Design pattern
Design patternDesign pattern
Design pattern
 
Decision tree
Decision treeDecision tree
Decision tree
 
Machine Learning part 3 - Introduction to data science
Machine Learning part 3 - Introduction to data science Machine Learning part 3 - Introduction to data science
Machine Learning part 3 - Introduction to data science
 
Classification of data using semi supervised learning a learning disability ...
Classification of data using semi supervised learning  a learning disability ...Classification of data using semi supervised learning  a learning disability ...
Classification of data using semi supervised learning a learning disability ...
 
A New Active Learning Technique Using Furthest Nearest Neighbour Criterion fo...
A New Active Learning Technique Using Furthest Nearest Neighbour Criterion fo...A New Active Learning Technique Using Furthest Nearest Neighbour Criterion fo...
A New Active Learning Technique Using Furthest Nearest Neighbour Criterion fo...
 
Been Kim - Interpretable machine learning, Nov 2015
Been Kim - Interpretable machine learning, Nov 2015Been Kim - Interpretable machine learning, Nov 2015
Been Kim - Interpretable machine learning, Nov 2015
 
IRJET- Facial Emotion Detection using Convolutional Neural Network
IRJET- Facial Emotion Detection using Convolutional Neural NetworkIRJET- Facial Emotion Detection using Convolutional Neural Network
IRJET- Facial Emotion Detection using Convolutional Neural Network
 
Data Clustering in Education for Students
Data Clustering in Education for StudentsData Clustering in Education for Students
Data Clustering in Education for Students
 
Predicting the Presence of Learning Motivation in Electronic Learning: A New ...
Predicting the Presence of Learning Motivation in Electronic Learning: A New ...Predicting the Presence of Learning Motivation in Electronic Learning: A New ...
Predicting the Presence of Learning Motivation in Electronic Learning: A New ...
 
Explainable AI in Healthcare
Explainable AI in HealthcareExplainable AI in Healthcare
Explainable AI in Healthcare
 
TS4-3: Takumi Sato from Nagoya Institute of Technology
TS4-3: Takumi Sato from Nagoya Institute of TechnologyTS4-3: Takumi Sato from Nagoya Institute of Technology
TS4-3: Takumi Sato from Nagoya Institute of Technology
 
Hybrid Classifier for Sentiment Analysis using Effective Pipelining
Hybrid Classifier for Sentiment Analysis using Effective PipeliningHybrid Classifier for Sentiment Analysis using Effective Pipelining
Hybrid Classifier for Sentiment Analysis using Effective Pipelining
 
Introduction to Interpretable Machine Learning
Introduction to Interpretable Machine LearningIntroduction to Interpretable Machine Learning
Introduction to Interpretable Machine Learning
 
Model Evaluation in the land of Deep Learning
Model Evaluation in the land of Deep LearningModel Evaluation in the land of Deep Learning
Model Evaluation in the land of Deep Learning
 

Similar to Theory and evaluation metrics for learning disentangled representations v2

Tech sem 2_dilip
Tech sem 2_dilipTech sem 2_dilip
Tech sem 2_dilip
Dilip Kolli
 
AI TESTING: ENSURING A GOOD DATA SPLIT BETWEEN DATA SETS (TRAINING AND TEST) ...
AI TESTING: ENSURING A GOOD DATA SPLIT BETWEEN DATA SETS (TRAINING AND TEST) ...AI TESTING: ENSURING A GOOD DATA SPLIT BETWEEN DATA SETS (TRAINING AND TEST) ...
AI TESTING: ENSURING A GOOD DATA SPLIT BETWEEN DATA SETS (TRAINING AND TEST) ...
ijsc
 
EXPLORATORY DATA ANALYSIS IN STATISTICAL MODeLING.pptx
EXPLORATORY DATA ANALYSIS IN STATISTICAL MODeLING.pptxEXPLORATORY DATA ANALYSIS IN STATISTICAL MODeLING.pptx
EXPLORATORY DATA ANALYSIS IN STATISTICAL MODeLING.pptx
rakeshreghu98
 
Introduction to Data Science - Week 3 - Steps involved in Data Science
Introduction to Data Science - Week 3 - Steps involved in Data ScienceIntroduction to Data Science - Week 3 - Steps involved in Data Science
Introduction to Data Science - Week 3 - Steps involved in Data Science
Ferdin Joe John Joseph PhD
 
Inverse Modeling for Cognitive Science "in the Wild"
Inverse Modeling for Cognitive Science "in the Wild"Inverse Modeling for Cognitive Science "in the Wild"
Inverse Modeling for Cognitive Science "in the Wild"
Aalto University
 
Hypothesis on Different Data Mining Algorithms
Hypothesis on Different Data Mining AlgorithmsHypothesis on Different Data Mining Algorithms
Hypothesis on Different Data Mining Algorithms
IJERA Editor
 
Construction of composite index: process & methods
Construction of composite index:  process & methodsConstruction of composite index:  process & methods
Construction of composite index: process & methods
gopichandbalusu
 
QUANTITATIVE-DATA.pptx
QUANTITATIVE-DATA.pptxQUANTITATIVE-DATA.pptx
QUANTITATIVE-DATA.pptx
ViaFortuna
 
Survey paper on Big Data Imputation and Privacy Algorithms
Survey paper on Big Data Imputation and Privacy AlgorithmsSurvey paper on Big Data Imputation and Privacy Algorithms
Survey paper on Big Data Imputation and Privacy Algorithms
IRJET Journal
 
Paper sharing_Explaining Data-Driven Decisions made by AI Systems_The Counter...
Paper sharing_Explaining Data-Driven Decisions made by AI Systems_The Counter...Paper sharing_Explaining Data-Driven Decisions made by AI Systems_The Counter...
Paper sharing_Explaining Data-Driven Decisions made by AI Systems_The Counter...
YOU SHENG CHEN
 
Explanations in Data Systems
Explanations in Data SystemsExplanations in Data Systems
Explanations in Data Systems
Fotis Savva
 
Data analysis for business and economics education
Data analysis for business and economics educationData analysis for business and economics education
Data analysis for business and economics education
FuleaAmena2
 
Hx3115011506
Hx3115011506Hx3115011506
Hx3115011506
IJERA Editor
 
لموعد الإثنين 03 يناير 2022 143 مبادرة #تواصل_تطوير المحاضرة ال 143 من المباد...
لموعد الإثنين 03 يناير 2022 143 مبادرة #تواصل_تطوير المحاضرة ال 143 من المباد...لموعد الإثنين 03 يناير 2022 143 مبادرة #تواصل_تطوير المحاضرة ال 143 من المباد...
لموعد الإثنين 03 يناير 2022 143 مبادرة #تواصل_تطوير المحاضرة ال 143 من المباد...
Egyptian Engineers Association
 
Enhancing a Social Science Model-building Workflow with Interactive Visualisa...
Enhancing a Social Science Model-building Workflow with Interactive Visualisa...Enhancing a Social Science Model-building Workflow with Interactive Visualisa...
Enhancing a Social Science Model-building Workflow with Interactive Visualisa...
Cagatay Turkay
 
​​Explainability in AI and Recommender systems: let’s make it interactive!
​​Explainability in AI and Recommender systems: let’s make it interactive!​​Explainability in AI and Recommender systems: let’s make it interactive!
​​Explainability in AI and Recommender systems: let’s make it interactive!
Eindhoven University of Technology / JADS
 
KatKennedy REU D.C. Poster
KatKennedy REU D.C. PosterKatKennedy REU D.C. Poster
KatKennedy REU D.C. Poster
Katlynn Kennedy
 
03_AJMS_298_21.pdf
03_AJMS_298_21.pdf03_AJMS_298_21.pdf
03_AJMS_298_21.pdf
BRNSS Publication Hub
 
Eliciting and Visualising Trust Expectations using Persona Trust Characterist...
Eliciting and Visualising Trust Expectations using Persona Trust Characterist...Eliciting and Visualising Trust Expectations using Persona Trust Characterist...
Eliciting and Visualising Trust Expectations using Persona Trust Characterist...
Shamal Faily
 
Qualitative data analysis - Student L
Qualitative data analysis - Student LQualitative data analysis - Student L
Qualitative data analysis - Student L
Lee Cox
 

Similar to Theory and evaluation metrics for learning disentangled representations v2 (20)

Tech sem 2_dilip
Tech sem 2_dilipTech sem 2_dilip
Tech sem 2_dilip
 
AI TESTING: ENSURING A GOOD DATA SPLIT BETWEEN DATA SETS (TRAINING AND TEST) ...
AI TESTING: ENSURING A GOOD DATA SPLIT BETWEEN DATA SETS (TRAINING AND TEST) ...AI TESTING: ENSURING A GOOD DATA SPLIT BETWEEN DATA SETS (TRAINING AND TEST) ...
AI TESTING: ENSURING A GOOD DATA SPLIT BETWEEN DATA SETS (TRAINING AND TEST) ...
 
EXPLORATORY DATA ANALYSIS IN STATISTICAL MODeLING.pptx
EXPLORATORY DATA ANALYSIS IN STATISTICAL MODeLING.pptxEXPLORATORY DATA ANALYSIS IN STATISTICAL MODeLING.pptx
EXPLORATORY DATA ANALYSIS IN STATISTICAL MODeLING.pptx
 
Introduction to Data Science - Week 3 - Steps involved in Data Science
Introduction to Data Science - Week 3 - Steps involved in Data ScienceIntroduction to Data Science - Week 3 - Steps involved in Data Science
Introduction to Data Science - Week 3 - Steps involved in Data Science
 
Inverse Modeling for Cognitive Science "in the Wild"
Inverse Modeling for Cognitive Science "in the Wild"Inverse Modeling for Cognitive Science "in the Wild"
Inverse Modeling for Cognitive Science "in the Wild"
 
Hypothesis on Different Data Mining Algorithms
Hypothesis on Different Data Mining AlgorithmsHypothesis on Different Data Mining Algorithms
Hypothesis on Different Data Mining Algorithms
 
Construction of composite index: process & methods
Construction of composite index:  process & methodsConstruction of composite index:  process & methods
Construction of composite index: process & methods
 
QUANTITATIVE-DATA.pptx
QUANTITATIVE-DATA.pptxQUANTITATIVE-DATA.pptx
QUANTITATIVE-DATA.pptx
 
Survey paper on Big Data Imputation and Privacy Algorithms
Survey paper on Big Data Imputation and Privacy AlgorithmsSurvey paper on Big Data Imputation and Privacy Algorithms
Survey paper on Big Data Imputation and Privacy Algorithms
 
Paper sharing_Explaining Data-Driven Decisions made by AI Systems_The Counter...
Paper sharing_Explaining Data-Driven Decisions made by AI Systems_The Counter...Paper sharing_Explaining Data-Driven Decisions made by AI Systems_The Counter...
Paper sharing_Explaining Data-Driven Decisions made by AI Systems_The Counter...
 
Explanations in Data Systems
Explanations in Data SystemsExplanations in Data Systems
Explanations in Data Systems
 
Data analysis for business and economics education
Data analysis for business and economics educationData analysis for business and economics education
Data analysis for business and economics education
 
Hx3115011506
Hx3115011506Hx3115011506
Hx3115011506
 
لموعد الإثنين 03 يناير 2022 143 مبادرة #تواصل_تطوير المحاضرة ال 143 من المباد...
لموعد الإثنين 03 يناير 2022 143 مبادرة #تواصل_تطوير المحاضرة ال 143 من المباد...لموعد الإثنين 03 يناير 2022 143 مبادرة #تواصل_تطوير المحاضرة ال 143 من المباد...
لموعد الإثنين 03 يناير 2022 143 مبادرة #تواصل_تطوير المحاضرة ال 143 من المباد...
 
Enhancing a Social Science Model-building Workflow with Interactive Visualisa...
Enhancing a Social Science Model-building Workflow with Interactive Visualisa...Enhancing a Social Science Model-building Workflow with Interactive Visualisa...
Enhancing a Social Science Model-building Workflow with Interactive Visualisa...
 
​​Explainability in AI and Recommender systems: let’s make it interactive!
​​Explainability in AI and Recommender systems: let’s make it interactive!​​Explainability in AI and Recommender systems: let’s make it interactive!
​​Explainability in AI and Recommender systems: let’s make it interactive!
 
KatKennedy REU D.C. Poster
KatKennedy REU D.C. PosterKatKennedy REU D.C. Poster
KatKennedy REU D.C. Poster
 
03_AJMS_298_21.pdf
03_AJMS_298_21.pdf03_AJMS_298_21.pdf
03_AJMS_298_21.pdf
 
Eliciting and Visualising Trust Expectations using Persona Trust Characterist...
Eliciting and Visualising Trust Expectations using Persona Trust Characterist...Eliciting and Visualising Trust Expectations using Persona Trust Characterist...
Eliciting and Visualising Trust Expectations using Persona Trust Characterist...
 
Qualitative data analysis - Student L
Qualitative data analysis - Student LQualitative data analysis - Student L
Qualitative data analysis - Student L
 

Recently uploaded

GNSS spoofing via SDR (Criptored Talks 2024)
GNSS spoofing via SDR (Criptored Talks 2024)GNSS spoofing via SDR (Criptored Talks 2024)
GNSS spoofing via SDR (Criptored Talks 2024)
Javier Junquera
 
Astute Business Solutions | Oracle Cloud Partner |
Astute Business Solutions | Oracle Cloud Partner |Astute Business Solutions | Oracle Cloud Partner |
Astute Business Solutions | Oracle Cloud Partner |
AstuteBusiness
 
"Scaling RAG Applications to serve millions of users", Kevin Goedecke
"Scaling RAG Applications to serve millions of users",  Kevin Goedecke"Scaling RAG Applications to serve millions of users",  Kevin Goedecke
"Scaling RAG Applications to serve millions of users", Kevin Goedecke
Fwdays
 
Harnessing the Power of NLP and Knowledge Graphs for Opioid Research
Harnessing the Power of NLP and Knowledge Graphs for Opioid ResearchHarnessing the Power of NLP and Knowledge Graphs for Opioid Research
Harnessing the Power of NLP and Knowledge Graphs for Opioid Research
Neo4j
 
QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...
QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...
QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...
AlexanderRichford
 
The Microsoft 365 Migration Tutorial For Beginner.pptx
The Microsoft 365 Migration Tutorial For Beginner.pptxThe Microsoft 365 Migration Tutorial For Beginner.pptx
The Microsoft 365 Migration Tutorial For Beginner.pptx
operationspcvita
 
Call Girls Chandigarh🔥7023059433🔥Agency Profile Escorts in Chandigarh Availab...
Call Girls Chandigarh🔥7023059433🔥Agency Profile Escorts in Chandigarh Availab...Call Girls Chandigarh🔥7023059433🔥Agency Profile Escorts in Chandigarh Availab...
Call Girls Chandigarh🔥7023059433🔥Agency Profile Escorts in Chandigarh Availab...
manji sharman06
 
Poznań ACE event - 19.06.2024 Team 24 Wrapup slidedeck
Poznań ACE event - 19.06.2024 Team 24 Wrapup slidedeckPoznań ACE event - 19.06.2024 Team 24 Wrapup slidedeck
Poznań ACE event - 19.06.2024 Team 24 Wrapup slidedeck
FilipTomaszewski5
 
AppSec PNW: Android and iOS Application Security with MobSF
AppSec PNW: Android and iOS Application Security with MobSFAppSec PNW: Android and iOS Application Security with MobSF
AppSec PNW: Android and iOS Application Security with MobSF
Ajin Abraham
 
Demystifying Knowledge Management through Storytelling
Demystifying Knowledge Management through StorytellingDemystifying Knowledge Management through Storytelling
Demystifying Knowledge Management through Storytelling
Enterprise Knowledge
 
AI in the Workplace Reskilling, Upskilling, and Future Work.pptx
AI in the Workplace Reskilling, Upskilling, and Future Work.pptxAI in the Workplace Reskilling, Upskilling, and Future Work.pptx
AI in the Workplace Reskilling, Upskilling, and Future Work.pptx
Sunil Jagani
 
JavaLand 2024: Application Development Green Masterplan
JavaLand 2024: Application Development Green MasterplanJavaLand 2024: Application Development Green Masterplan
JavaLand 2024: Application Development Green Masterplan
Miro Wengner
 
What is an RPA CoE? Session 1 – CoE Vision
What is an RPA CoE?  Session 1 – CoE VisionWhat is an RPA CoE?  Session 1 – CoE Vision
What is an RPA CoE? Session 1 – CoE Vision
DianaGray10
 
PRODUCT LISTING OPTIMIZATION PRESENTATION.pptx
PRODUCT LISTING OPTIMIZATION PRESENTATION.pptxPRODUCT LISTING OPTIMIZATION PRESENTATION.pptx
PRODUCT LISTING OPTIMIZATION PRESENTATION.pptx
christinelarrosa
 
Northern Engraving | Modern Metal Trim, Nameplates and Appliance Panels
Northern Engraving | Modern Metal Trim, Nameplates and Appliance PanelsNorthern Engraving | Modern Metal Trim, Nameplates and Appliance Panels
Northern Engraving | Modern Metal Trim, Nameplates and Appliance Panels
Northern Engraving
 
Christine's Product Research Presentation.pptx
Christine's Product Research Presentation.pptxChristine's Product Research Presentation.pptx
Christine's Product Research Presentation.pptx
christinelarrosa
 
Dandelion Hashtable: beyond billion requests per second on a commodity server
Dandelion Hashtable: beyond billion requests per second on a commodity serverDandelion Hashtable: beyond billion requests per second on a commodity server
Dandelion Hashtable: beyond billion requests per second on a commodity server
Antonios Katsarakis
 
A Deep Dive into ScyllaDB's Architecture
A Deep Dive into ScyllaDB's ArchitectureA Deep Dive into ScyllaDB's Architecture
A Deep Dive into ScyllaDB's Architecture
ScyllaDB
 
Leveraging the Graph for Clinical Trials and Standards
Leveraging the Graph for Clinical Trials and StandardsLeveraging the Graph for Clinical Trials and Standards
Leveraging the Graph for Clinical Trials and Standards
Neo4j
 
"What does it really mean for your system to be available, or how to define w...
"What does it really mean for your system to be available, or how to define w..."What does it really mean for your system to be available, or how to define w...
"What does it really mean for your system to be available, or how to define w...
Fwdays
 

Recently uploaded (20)

GNSS spoofing via SDR (Criptored Talks 2024)
GNSS spoofing via SDR (Criptored Talks 2024)GNSS spoofing via SDR (Criptored Talks 2024)
GNSS spoofing via SDR (Criptored Talks 2024)
 
Astute Business Solutions | Oracle Cloud Partner |
Astute Business Solutions | Oracle Cloud Partner |Astute Business Solutions | Oracle Cloud Partner |
Astute Business Solutions | Oracle Cloud Partner |
 
"Scaling RAG Applications to serve millions of users", Kevin Goedecke
"Scaling RAG Applications to serve millions of users",  Kevin Goedecke"Scaling RAG Applications to serve millions of users",  Kevin Goedecke
"Scaling RAG Applications to serve millions of users", Kevin Goedecke
 
Harnessing the Power of NLP and Knowledge Graphs for Opioid Research
Harnessing the Power of NLP and Knowledge Graphs for Opioid ResearchHarnessing the Power of NLP and Knowledge Graphs for Opioid Research
Harnessing the Power of NLP and Knowledge Graphs for Opioid Research
 
QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...
QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...
QR Secure: A Hybrid Approach Using Machine Learning and Security Validation F...
 
The Microsoft 365 Migration Tutorial For Beginner.pptx
The Microsoft 365 Migration Tutorial For Beginner.pptxThe Microsoft 365 Migration Tutorial For Beginner.pptx
The Microsoft 365 Migration Tutorial For Beginner.pptx
 
Call Girls Chandigarh🔥7023059433🔥Agency Profile Escorts in Chandigarh Availab...
Call Girls Chandigarh🔥7023059433🔥Agency Profile Escorts in Chandigarh Availab...Call Girls Chandigarh🔥7023059433🔥Agency Profile Escorts in Chandigarh Availab...
Call Girls Chandigarh🔥7023059433🔥Agency Profile Escorts in Chandigarh Availab...
 
Poznań ACE event - 19.06.2024 Team 24 Wrapup slidedeck
Poznań ACE event - 19.06.2024 Team 24 Wrapup slidedeckPoznań ACE event - 19.06.2024 Team 24 Wrapup slidedeck
Poznań ACE event - 19.06.2024 Team 24 Wrapup slidedeck
 
AppSec PNW: Android and iOS Application Security with MobSF
AppSec PNW: Android and iOS Application Security with MobSFAppSec PNW: Android and iOS Application Security with MobSF
AppSec PNW: Android and iOS Application Security with MobSF
 
Demystifying Knowledge Management through Storytelling
Demystifying Knowledge Management through StorytellingDemystifying Knowledge Management through Storytelling
Demystifying Knowledge Management through Storytelling
 
AI in the Workplace Reskilling, Upskilling, and Future Work.pptx
AI in the Workplace Reskilling, Upskilling, and Future Work.pptxAI in the Workplace Reskilling, Upskilling, and Future Work.pptx
AI in the Workplace Reskilling, Upskilling, and Future Work.pptx
 
JavaLand 2024: Application Development Green Masterplan
JavaLand 2024: Application Development Green MasterplanJavaLand 2024: Application Development Green Masterplan
JavaLand 2024: Application Development Green Masterplan
 
What is an RPA CoE? Session 1 – CoE Vision
What is an RPA CoE?  Session 1 – CoE VisionWhat is an RPA CoE?  Session 1 – CoE Vision
What is an RPA CoE? Session 1 – CoE Vision
 
PRODUCT LISTING OPTIMIZATION PRESENTATION.pptx
PRODUCT LISTING OPTIMIZATION PRESENTATION.pptxPRODUCT LISTING OPTIMIZATION PRESENTATION.pptx
PRODUCT LISTING OPTIMIZATION PRESENTATION.pptx
 
Northern Engraving | Modern Metal Trim, Nameplates and Appliance Panels
Northern Engraving | Modern Metal Trim, Nameplates and Appliance PanelsNorthern Engraving | Modern Metal Trim, Nameplates and Appliance Panels
Northern Engraving | Modern Metal Trim, Nameplates and Appliance Panels
 
Christine's Product Research Presentation.pptx
Christine's Product Research Presentation.pptxChristine's Product Research Presentation.pptx
Christine's Product Research Presentation.pptx
 
Dandelion Hashtable: beyond billion requests per second on a commodity server
Dandelion Hashtable: beyond billion requests per second on a commodity serverDandelion Hashtable: beyond billion requests per second on a commodity server
Dandelion Hashtable: beyond billion requests per second on a commodity server
 
A Deep Dive into ScyllaDB's Architecture
A Deep Dive into ScyllaDB's ArchitectureA Deep Dive into ScyllaDB's Architecture
A Deep Dive into ScyllaDB's Architecture
 
Leveraging the Graph for Clinical Trials and Standards
Leveraging the Graph for Clinical Trials and StandardsLeveraging the Graph for Clinical Trials and Standards
Leveraging the Graph for Clinical Trials and Standards
 
"What does it really mean for your system to be available, or how to define w...
"What does it really mean for your system to be available, or how to define w..."What does it really mean for your system to be available, or how to define w...
"What does it really mean for your system to be available, or how to define w...
 

Theory and evaluation metrics for learning disentangled representations v2

  • 1. Theory and Evaluation Metrics for Learning Disentangled Representations Kien Do and Truyen Tran Applied AI Institute (A2I2), Deakin University, Australia 12020
  • 2. Examples about disentangled representations 2 Overcoming the Disentanglement vs Reconstruction Trade-off via Jacobian Supervision (Lezama et. al., ICLR-2019)
  • 3. Current challenges • The mathematic interpretation of disentangled representations is still not clear • Most evaluations metrics are heuristic and not robust 3
  • 4. Our goals • Provide a mathematically formal definition of disentangled representations • Propose proper, theoretically driven metrics for fair evaluations 4
  • 5. An informal definition Disentanglement: A process of decorrelating data into separate pieces of information, each of which corresponds to a predefined concept. This suggests 3 main properties of disentangled representations: • Informativeness: Large enough pieces of information • Separability: Pieces of information are not overlapped • Interpretability: Each piece of information corresponds to a concept 6
  • 6. Mathematic formulas of these properties • The informativeness of is defined by: • are separable w.r.t. if: • are fully separable (independent) if: • is fully interpretable w.r.t. if: 7
  • 7. Definition of disentangled representations 8
  • 8. Compared to other related works • Higgins et. al., 2018 [1] attempted to define disentanglement based on group theory. Their definition shares similar observations as ours but is less convenient for probabilistic model design. • Eastwood et. al., 2018 [2] did NOT provide explicit definition of disentangled representations but characterized them along 3 dimensions: disentanglement, compactness, and informativeness. • The main drawback of disentanglement and compactness is that they require knowledge of ALL ground truth factors of variation. 9 [1] Towards a definition of disentangled representations (Higgins et. al., ICLR-2018) [2] A framework for the quantitative evaluation of disentangled representations (Eastwood et. al., ICLR-2018)
  • 9. Our proposed metrics • Informativeness: • Separability + Informativeness (no labeled factors available): 10 where
  • 10. Our proposed metrics (cont.) • Separability + Interpretability (labeled factors provided): where and is the highest and second highest MI values between every and . 11
  • 11. Advantages of our metrics • Support both supervised/unsupervised models • Can be applied for real datasets • Do not require any training procedure • Provide consistent results across different models • Agree with visual results 12
  • 12. Experiments and Results Please refer to our paper! 13
  • 13. 14 Thank you for listening!