SlideShare a Scribd company logo
1 of 53
Download to read offline
CS592 Presentation #18
Composing graphical models with
neural networks for structured
representations and fast inference
20173586 Jeongmin Cha
20184666 Yajie Zhou
20174463 Jaesung Choe
Content
1. Motivation
2. Modeling idea
3. Structural Variational Autoencoder (SVAE)
4. Background
5. Main algorithm
6. Experiment
7. Group Discussion Point
1. Motivation
● How can we build interpretable models of high-dimensional data?
● modeling video of a mouse
● a mouse usually repeats a certain behavior
dart groom rear
1. Motivation
● We want a model
○ can explain which behavior
most is performing at each frame
1. Motivation
● What we want to do is ...
● segment and categorize mouse behavior from the video
● Q: generative vs discriminative model for this task?
1. Motivation
● What we want to do is ...
● segment and categorize mouse behavior from the video
● Q: generative vs discriminative model for this task?
○ We can use both
○ a large number of labeled data needed in discriminative scheme
○ discriminative model relaxes conditional independence assumption
so may achieve better predictive result
1. Motivation
● What we want to do is ...
● segment and categorize mouse behavior from the video
● Q: generative vs discriminative model for this task?
○ We can use both
○ a large number of unlabeled data from a small number of labeled data
1. Motivation
● What we want to do is ...
● segment and categorize mouse behavior from the video
● Q: generative vs discriminative model for this task?
○ We can use both
○ a large number of unlabeled data from a small number of labeled data
○ This paper want to build a generative model for video of a mouse
1. Motivation
● a generative model for video of a mouse
● a mouse repeats certain behaviors
GMM = one solution
1. Motivation
● However, what if the data is not well described by Gaussian?
1. Motivation
● a mixture of gaussians fits the data poorly
● reports too many clusters (not natural clustering result)
GMMGMM
1. Motivation
● neural network fits data well
● but, difficult to interpret in high dimensions (lack interpretability)
GMM
density net
(VAE)
1. Motivation
● neural network fits data well
● but, difficult to interpret in high dimensions (lack interpretability)
● does not explicitly represent discrete mixture components,
GMM
density net
(VAE)
An appropriate model might
switch between discrete states
1. Motivation
● How about combining both? (Graphical model + Deep Learning)
● Structured Variational AutoEncoder (SVAE)
density net
(VAE)
GMM GMM SVAE
1. Motivation
● Q: Graphical model vs Deep Learning, pros and cons?
1. Motivation
● Q: Graphical model vs Deep Learning, pros and cons?
● specify explicit relationship between variables before learning
○ Graphical model configuration starts from a higher level (deduction)
○ Deep learning configuration starts from a lower level (induction)
1. Motivation
● Graphical model
○ + interpretable, structured
representations
○ + data and computational efficiency
○ - strong assumptions may not fit
○ - feature engineering
○ - top-down inference
● Deep learning
○ - not directly interpretable structure
○ - can require lots of data
○ + flexible representations, learn
automatically
○ + feature learning
○ + recognition networks (bottom-up)
2. Modeling idea
● graphical models on latent variables
○ structured probability distributions
○ fast exact inference subroutines
● neural network models (VAE) for observations
○ produce flexible non-linear feature manifold
■ nonlinear high-dimensional data to low-dimensional and dense representations
○ recognition network
■ instead of learning variational distribution parameters directly
■ map observations to conjugate graphical model potentials
■
3. Structure Variational AutoEncoder (SVAE)
Under the exponential conjugate property we can define SAVE as below,
where p(θ) is the prior distribution and the p(x|θ) is the posterior distribution.
Statistics function is defined and the partition function is
described as,
Finally, we should like to infer the likelihood.
3. Structure Variational AutoEncoder (SVAE)
Under the exponential conjugate property we can define SAVE as below,
where p(θ) is the prior distribution and the p(x|θ) is the posterior distribution.
Statistics function is defined and the partition function is
described as,
Finally, we should like to infer the likelihood.
Discussion Point:
Can you tell the fundamental
difference between VAE and
SAVE ?
3. Structure Variational AutoEncoder (SVAE)
Under the exponential conjugate property we can define SAVE as below,
where p(θ) is the prior distribution and the p(x|θ) is the posterior distribution.
Statistics function is defined and the partition function is
described as,
Finally, we should like to infer the likelihood.
Discussion Point:
Can you tell the fundamental
difference between VAE and
SAVE ? : Conjugate property
4. Background : conjugate distribution (VAE vs SVAE)
if the posterior distributions p(x|θ) are in the same probability distribution family as the prior probability
distribution p(θ), the prior and posterior are then called conjugate distributions.
What is conjugate distribution?
if the likelihood function is Poisson distribution, choosing a Poisson prior over the parameter λ will ensure that the
posterior distribution is also Poisson distribution.
Example
where λ = 4 where k = 4 and θ = 1
Posteriori : Prior:
4. Background : conjugate distribution.
if the posterior distributions p(x|θ) are in the same probability distribution family as the prior probability
distribution p(θ), the prior and posterior are then called conjugate distributions.
What is conjugate distribution?
if the likelihood function is Poisson distribution, choosing a Poisson prior over the parameter λ will ensure that the
posterior distribution is also Poisson distribution.
Example
Likelihood:
(assume i=1, …, 6)
If we set k = 10 and θ = 0.5, Prior becomes posterior:
Prior:
4. Background : conjugate distribution.
if the posterior distributions p(x|θ) are in the same probability distribution family as the prior probability
distribution p(θ), the prior and posterior are then called conjugate distributions.
What is conjugate distribution?
estimate the likelihood by updating the parameters of our prior
 —  reflecting a new mean and confidence level
4. Background : conjugate distribution.
if the posterior distributions p(x|θ) are in the same probability distribution family as the prior probability
distribution p(θ), the prior and posterior are then called conjugate distributions.
What is conjugate distribution?
estimate the likelihood by updating the parameters of our prior
 —  reflecting a new mean and confidence level
Discussion Point:
Why the upper property is
important in SVAE ?
4. Background : conjugate distribution.
if the posterior distributions p(θ | x) are in the same probability distribution family as the prior probability
distribution p(θ), the prior and posterior are then called conjugate distributions.
What is conjugate distribution?
estimate the likelihood by updating the parameters of our prior
 —  reflecting a new mean and confidence level
Discussion Point:
Why the upper property is
important in SVAE ?
A : Conjugacy property is
useful in Bayesian inference !
4. Background : conjugate distribution.
the integral of the marginal likelihood = is intractable.
Intractability in VAE.
Conjugacy in SVAE (Proposition B.4)
=
where the posterior p(θ|x) is in the same exponential family as p(θ) with the natural parameter
=
, and d are statistic function.
4. Background : conjugate distribution.
the integral of the marginal likelihood = is intractable.
Intractability in VAE.
Conjugacy in SVAE (Proposition B.4)
=
where the posterior p(θ|x) is in the same exponential family as p(θ) with the natural parameter
=
, and d are statistic function.
VAE handles a general non-conjugate observation models by introducing recognition network.
4. Background : conjugate distribution.
the integral of the marginal likelihood = is intractable.
Intractability in VAE.
Conjugacy in SVAE (Proposition B.4)
where the posterior p(θ|x) is in the same exponential family as p(θ)
=
, and are statistic function.
4. Background : conjugate distribution.
the integral of the marginal likelihood = is intractable.
Intractability in VAE.
Conjugacy in SVAE (Proposition B.4)
=
where the posterior p(θ|x) is in the same exponential family as p(θ) with the natural parameter
=
, and d are statistic function.
4. Background : conjugate distribution.
the integral of the marginal likelihood = is intractable.
Intractability in VAE.
Conjugacy in SVAE (Proposition B.4)
= =
=
This relationship is useful in Bayesian inference under the conjugacy property.
5.1 Conjugate Inference
Inference
5.1 Conjugate Inference
5.1 Conjugate Inference
5.2 Non-conjugate Inference
Discussion Point:
is still linear Gaussian?
what is conjugate prior?
5.2 Non-conjugate Inference
SVAEs: recognition networks output conjugate potentials, then apply fast
graphical model inference
5.2 Non-conjugate Inference
5.2 Non-conjugate Inference
Discussion Point:
How to optimize this bound?
5.2 Non-conjugate Inference
6. Experiments
6. Experiments
6. Experiments
https://www.youtube.com/watch?v=9WSb-89UsEo&t=60s
(This video is disabled to be watched on other sites except for Youtube)
7. Group discussion
Group Discussion Point:
VAE vs SVAE : which model can have better
performance? (Is it strike or ball?)
: Strike
: None
: Strike
: Ball
፠ Supplementary material
- For those who are not familiar with the baseball rules.
https://www.youtube.com/watch?v=9WSb-89UsEo&t=60s
(This video is disabled to be watched on other sites except for Youtube)
7. Group discussion
Group Discussion Point:
VAE vs SVAE : which model can have better
performance? (Is it strike or ball?)
: Strike
: None
: Strike
: Ball
፠ Supplementary material
- For those who are not familiar with the baseball rules.
Hint or not
If SVAE follows the linear-chain structure,
(expect) SVAE can predict better accuracy in video classification.
(expect) VAE would be better for the single image classification.
7. Group discussion
If SVAE follows the linear-chain structure,
(expect) SVAE can predict better accuracy in video classification.
(expect) VAE would be better for the single image classification.
7. Group discussion
NO
By the way, what is the result?
strike - ball
7. Group discussion
By the way, what is the result?
strike - ball
7. Group discussion
scoreboard
By the way, what is the result? Strick!!
7. Group discussion
strike - ball
scoreboard
By the way, what is the result? Strick!! How did you check the results?
7. Group discussion
strike - ball
scoreboard
By the way, what is the result? Strick!! How did you check the results?
I think just single frame is enough !!
7. Group discussion
strike - ball
scoreboard
As we check the scoreboard,
AI also look at the scoreboard for the inference.
In other words, we do not need sequential frames.
(Our expectation) VAE would be better.
If we mask the scoreboard,
(our expectation) SVAE would be better.
7. Group discussion
Where AI is looking at (i.e. high attention)
Mask(non-observable area)
Thank you

More Related Content

What's hot

(研究会輪読) Facial Landmark Detection by Deep Multi-task Learning
(研究会輪読) Facial Landmark Detection by Deep Multi-task Learning(研究会輪読) Facial Landmark Detection by Deep Multi-task Learning
(研究会輪読) Facial Landmark Detection by Deep Multi-task LearningMasahiro Suzuki
 
Probability distributions for ml
Probability distributions for mlProbability distributions for ml
Probability distributions for mlSung Yub Kim
 
An introduction to neural network
An introduction to neural networkAn introduction to neural network
An introduction to neural networktuxette
 
RECENT ADVANCES in PREDICTIVE (MACHINE) LEARNING
RECENT ADVANCES in PREDICTIVE (MACHINE) LEARNINGRECENT ADVANCES in PREDICTIVE (MACHINE) LEARNING
RECENT ADVANCES in PREDICTIVE (MACHINE) LEARNINGbutest
 
【博士論文発表会】パラメータ制約付き特異モデルの統計的学習理論
【博士論文発表会】パラメータ制約付き特異モデルの統計的学習理論【博士論文発表会】パラメータ制約付き特異モデルの統計的学習理論
【博士論文発表会】パラメータ制約付き特異モデルの統計的学習理論Naoki Hayashi
 
An introduction to neural networks
An introduction to neural networksAn introduction to neural networks
An introduction to neural networkstuxette
 
MachineLearning.ppt
MachineLearning.pptMachineLearning.ppt
MachineLearning.pptbutest
 
Ch 9-1.Machine Learning: Symbol-based
Ch 9-1.Machine Learning: Symbol-basedCh 9-1.Machine Learning: Symbol-based
Ch 9-1.Machine Learning: Symbol-basedbutest
 
IEEESSCI2017-FOCI4-1039
IEEESSCI2017-FOCI4-1039IEEESSCI2017-FOCI4-1039
IEEESSCI2017-FOCI4-1039Naoki Hayashi
 
25 Machine Learning Unsupervised Learaning K-means K-centers
25 Machine Learning Unsupervised Learaning K-means K-centers25 Machine Learning Unsupervised Learaning K-means K-centers
25 Machine Learning Unsupervised Learaning K-means K-centersAndres Mendez-Vazquez
 
Presentation on Machine Learning and Data Mining
Presentation on Machine Learning and Data MiningPresentation on Machine Learning and Data Mining
Presentation on Machine Learning and Data Miningbutest
 
GAN(と強化学習との関係)
GAN(と強化学習との関係)GAN(と強化学習との関係)
GAN(と強化学習との関係)Masahiro Suzuki
 
Foundations of Machine Learning
Foundations of Machine LearningFoundations of Machine Learning
Foundations of Machine Learningmahutte
 
[DL輪読会]DisCo RL: Distribution-Conditioned Reinforcement Learning for General...
[DL輪読会]DisCo RL:  Distribution-Conditioned Reinforcement Learning for General...[DL輪読会]DisCo RL:  Distribution-Conditioned Reinforcement Learning for General...
[DL輪読会]DisCo RL: Distribution-Conditioned Reinforcement Learning for General...Deep Learning JP
 

What's hot (20)

Chapter 5 (final)
Chapter 5 (final)Chapter 5 (final)
Chapter 5 (final)
 
(研究会輪読) Facial Landmark Detection by Deep Multi-task Learning
(研究会輪読) Facial Landmark Detection by Deep Multi-task Learning(研究会輪読) Facial Landmark Detection by Deep Multi-task Learning
(研究会輪読) Facial Landmark Detection by Deep Multi-task Learning
 
Probability distributions for ml
Probability distributions for mlProbability distributions for ml
Probability distributions for ml
 
An introduction to neural network
An introduction to neural networkAn introduction to neural network
An introduction to neural network
 
RECENT ADVANCES in PREDICTIVE (MACHINE) LEARNING
RECENT ADVANCES in PREDICTIVE (MACHINE) LEARNINGRECENT ADVANCES in PREDICTIVE (MACHINE) LEARNING
RECENT ADVANCES in PREDICTIVE (MACHINE) LEARNING
 
【博士論文発表会】パラメータ制約付き特異モデルの統計的学習理論
【博士論文発表会】パラメータ制約付き特異モデルの統計的学習理論【博士論文発表会】パラメータ制約付き特異モデルの統計的学習理論
【博士論文発表会】パラメータ制約付き特異モデルの統計的学習理論
 
An introduction to neural networks
An introduction to neural networksAn introduction to neural networks
An introduction to neural networks
 
MachineLearning.ppt
MachineLearning.pptMachineLearning.ppt
MachineLearning.ppt
 
Chapter 9
Chapter 9Chapter 9
Chapter 9
 
06 Machine Learning - Naive Bayes
06 Machine Learning - Naive Bayes06 Machine Learning - Naive Bayes
06 Machine Learning - Naive Bayes
 
AI Lesson 34
AI Lesson 34AI Lesson 34
AI Lesson 34
 
Ch 9-1.Machine Learning: Symbol-based
Ch 9-1.Machine Learning: Symbol-basedCh 9-1.Machine Learning: Symbol-based
Ch 9-1.Machine Learning: Symbol-based
 
IEEESSCI2017-FOCI4-1039
IEEESSCI2017-FOCI4-1039IEEESSCI2017-FOCI4-1039
IEEESSCI2017-FOCI4-1039
 
25 Machine Learning Unsupervised Learaning K-means K-centers
25 Machine Learning Unsupervised Learaning K-means K-centers25 Machine Learning Unsupervised Learaning K-means K-centers
25 Machine Learning Unsupervised Learaning K-means K-centers
 
Presentation on Machine Learning and Data Mining
Presentation on Machine Learning and Data MiningPresentation on Machine Learning and Data Mining
Presentation on Machine Learning and Data Mining
 
Ai inductive bias and knowledge
Ai inductive bias and knowledgeAi inductive bias and knowledge
Ai inductive bias and knowledge
 
GAN(と強化学習との関係)
GAN(と強化学習との関係)GAN(と強化学習との関係)
GAN(と強化学習との関係)
 
Foundations of Machine Learning
Foundations of Machine LearningFoundations of Machine Learning
Foundations of Machine Learning
 
F0422052058
F0422052058F0422052058
F0422052058
 
[DL輪読会]DisCo RL: Distribution-Conditioned Reinforcement Learning for General...
[DL輪読会]DisCo RL:  Distribution-Conditioned Reinforcement Learning for General...[DL輪読会]DisCo RL:  Distribution-Conditioned Reinforcement Learning for General...
[DL輪読会]DisCo RL: Distribution-Conditioned Reinforcement Learning for General...
 

Similar to Composing graphical models with neural networks for structured representations and fast inference

Neural Nets Deconstructed
Neural Nets DeconstructedNeural Nets Deconstructed
Neural Nets DeconstructedPaul Sterk
 
17 Machine Learning Radial Basis Functions
17 Machine Learning Radial Basis Functions17 Machine Learning Radial Basis Functions
17 Machine Learning Radial Basis FunctionsAndres Mendez-Vazquez
 
Modeling uncertainty in deep learning
Modeling uncertainty in deep learning Modeling uncertainty in deep learning
Modeling uncertainty in deep learning Sungjoon Choi
 
"Automatic Variational Inference in Stan" NIPS2015_yomi2016-01-20
"Automatic Variational Inference in Stan" NIPS2015_yomi2016-01-20"Automatic Variational Inference in Stan" NIPS2015_yomi2016-01-20
"Automatic Variational Inference in Stan" NIPS2015_yomi2016-01-20Yuta Kashino
 
Learning group variational inference
Learning group  variational inferenceLearning group  variational inference
Learning group variational inferenceShuai Zhang
 
[DL輪読会]Generative Models of Visually Grounded Imagination
[DL輪読会]Generative Models of Visually Grounded Imagination[DL輪読会]Generative Models of Visually Grounded Imagination
[DL輪読会]Generative Models of Visually Grounded ImaginationDeep Learning JP
 
Computational Biology, Part 4 Protein Coding Regions
Computational Biology, Part 4 Protein Coding RegionsComputational Biology, Part 4 Protein Coding Regions
Computational Biology, Part 4 Protein Coding Regionsbutest
 
PRML Chapter 5
PRML Chapter 5PRML Chapter 5
PRML Chapter 5Sunwoo Kim
 
Inductive bias
Inductive biasInductive bias
Inductive biasswapnac12
 
ML-Inductive Bias.pdf
ML-Inductive Bias.pdfML-Inductive Bias.pdf
ML-Inductive Bias.pdfShivaVemula2
 
SVM - Functional Verification
SVM - Functional VerificationSVM - Functional Verification
SVM - Functional VerificationSai Kiran Kadam
 
Fuzzy logic and fuzzy time series edited
Fuzzy logic and fuzzy time series   editedFuzzy logic and fuzzy time series   edited
Fuzzy logic and fuzzy time series editedProf Dr S.M.Aqil Burney
 
MachineLearning.ppt
MachineLearning.pptMachineLearning.ppt
MachineLearning.pptbutest
 
MachineLearning.ppt
MachineLearning.pptMachineLearning.ppt
MachineLearning.pptbutest
 
OBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkk
OBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkkOBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkk
OBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkkshesnasuneer
 
OBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkk
OBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkkOBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkk
OBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkkshesnasuneer
 
Joint contrastive learning with infinite possibilities
Joint contrastive learning with infinite possibilitiesJoint contrastive learning with infinite possibilities
Joint contrastive learning with infinite possibilitiestaeseon ryu
 

Similar to Composing graphical models with neural networks for structured representations and fast inference (20)

Neural Nets Deconstructed
Neural Nets DeconstructedNeural Nets Deconstructed
Neural Nets Deconstructed
 
17 Machine Learning Radial Basis Functions
17 Machine Learning Radial Basis Functions17 Machine Learning Radial Basis Functions
17 Machine Learning Radial Basis Functions
 
Modeling uncertainty in deep learning
Modeling uncertainty in deep learning Modeling uncertainty in deep learning
Modeling uncertainty in deep learning
 
"Automatic Variational Inference in Stan" NIPS2015_yomi2016-01-20
"Automatic Variational Inference in Stan" NIPS2015_yomi2016-01-20"Automatic Variational Inference in Stan" NIPS2015_yomi2016-01-20
"Automatic Variational Inference in Stan" NIPS2015_yomi2016-01-20
 
Fuzzy sets
Fuzzy sets Fuzzy sets
Fuzzy sets
 
G
GG
G
 
Learning group variational inference
Learning group  variational inferenceLearning group  variational inference
Learning group variational inference
 
[DL輪読会]Generative Models of Visually Grounded Imagination
[DL輪読会]Generative Models of Visually Grounded Imagination[DL輪読会]Generative Models of Visually Grounded Imagination
[DL輪読会]Generative Models of Visually Grounded Imagination
 
Computational Biology, Part 4 Protein Coding Regions
Computational Biology, Part 4 Protein Coding RegionsComputational Biology, Part 4 Protein Coding Regions
Computational Biology, Part 4 Protein Coding Regions
 
PRML Chapter 5
PRML Chapter 5PRML Chapter 5
PRML Chapter 5
 
Inductive bias
Inductive biasInductive bias
Inductive bias
 
ML-Inductive Bias.pdf
ML-Inductive Bias.pdfML-Inductive Bias.pdf
ML-Inductive Bias.pdf
 
SVM - Functional Verification
SVM - Functional VerificationSVM - Functional Verification
SVM - Functional Verification
 
Fuzzy logic and fuzzy time series edited
Fuzzy logic and fuzzy time series   editedFuzzy logic and fuzzy time series   edited
Fuzzy logic and fuzzy time series edited
 
MachineLearning.ppt
MachineLearning.pptMachineLearning.ppt
MachineLearning.ppt
 
MachineLearning.ppt
MachineLearning.pptMachineLearning.ppt
MachineLearning.ppt
 
OBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkk
OBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkkOBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkk
OBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkk
 
OBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkk
OBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkkOBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkk
OBJECTRECOGNITION1.pptxjjjkkkkjjjjkkkkkkk
 
Joint contrastive learning with infinite possibilities
Joint contrastive learning with infinite possibilitiesJoint contrastive learning with infinite possibilities
Joint contrastive learning with infinite possibilities
 
Quantum Deep Learning
Quantum Deep LearningQuantum Deep Learning
Quantum Deep Learning
 

More from Jeongmin Cha

차정민 (소프트웨어 엔지니어) 이력서 + 경력기술서
차정민 (소프트웨어 엔지니어) 이력서 + 경력기술서차정민 (소프트웨어 엔지니어) 이력서 + 경력기술서
차정민 (소프트웨어 엔지니어) 이력서 + 경력기술서Jeongmin Cha
 
Causal Effect Inference with Deep Latent-Variable Models
Causal Effect Inference with Deep Latent-Variable ModelsCausal Effect Inference with Deep Latent-Variable Models
Causal Effect Inference with Deep Latent-Variable ModelsJeongmin Cha
 
Sparse Additive Models (SPAM)
Sparse Additive Models (SPAM)Sparse Additive Models (SPAM)
Sparse Additive Models (SPAM)Jeongmin Cha
 
Waterful Application (iOS + AppleWatch)
Waterful Application (iOS + AppleWatch)Waterful Application (iOS + AppleWatch)
Waterful Application (iOS + AppleWatch)Jeongmin Cha
 
시스템 프로그램 설계 2 최종발표 (차정민, 조경재)
시스템 프로그램 설계 2 최종발표 (차정민, 조경재)시스템 프로그램 설계 2 최종발표 (차정민, 조경재)
시스템 프로그램 설계 2 최종발표 (차정민, 조경재)Jeongmin Cha
 
시스템 프로그램 설계1 최종발표
시스템 프로그램 설계1 최종발표시스템 프로그램 설계1 최종발표
시스템 프로그램 설계1 최종발표Jeongmin Cha
 
마이크로프로세서 응용(2013-2)
마이크로프로세서 응용(2013-2)마이크로프로세서 응용(2013-2)
마이크로프로세서 응용(2013-2)Jeongmin Cha
 
최종발표
최종발표최종발표
최종발표Jeongmin Cha
 

More from Jeongmin Cha (8)

차정민 (소프트웨어 엔지니어) 이력서 + 경력기술서
차정민 (소프트웨어 엔지니어) 이력서 + 경력기술서차정민 (소프트웨어 엔지니어) 이력서 + 경력기술서
차정민 (소프트웨어 엔지니어) 이력서 + 경력기술서
 
Causal Effect Inference with Deep Latent-Variable Models
Causal Effect Inference with Deep Latent-Variable ModelsCausal Effect Inference with Deep Latent-Variable Models
Causal Effect Inference with Deep Latent-Variable Models
 
Sparse Additive Models (SPAM)
Sparse Additive Models (SPAM)Sparse Additive Models (SPAM)
Sparse Additive Models (SPAM)
 
Waterful Application (iOS + AppleWatch)
Waterful Application (iOS + AppleWatch)Waterful Application (iOS + AppleWatch)
Waterful Application (iOS + AppleWatch)
 
시스템 프로그램 설계 2 최종발표 (차정민, 조경재)
시스템 프로그램 설계 2 최종발표 (차정민, 조경재)시스템 프로그램 설계 2 최종발표 (차정민, 조경재)
시스템 프로그램 설계 2 최종발표 (차정민, 조경재)
 
시스템 프로그램 설계1 최종발표
시스템 프로그램 설계1 최종발표시스템 프로그램 설계1 최종발표
시스템 프로그램 설계1 최종발표
 
마이크로프로세서 응용(2013-2)
마이크로프로세서 응용(2013-2)마이크로프로세서 응용(2013-2)
마이크로프로세서 응용(2013-2)
 
최종발표
최종발표최종발표
최종발표
 

Recently uploaded

Simplified FDO Manufacturing Flow with TPMs _ Liam at Infineon.pdf
Simplified FDO Manufacturing Flow with TPMs _ Liam at Infineon.pdfSimplified FDO Manufacturing Flow with TPMs _ Liam at Infineon.pdf
Simplified FDO Manufacturing Flow with TPMs _ Liam at Infineon.pdfFIDO Alliance
 
The Value of Certifying Products for FDO _ Paul at FIDO Alliance.pdf
The Value of Certifying Products for FDO _ Paul at FIDO Alliance.pdfThe Value of Certifying Products for FDO _ Paul at FIDO Alliance.pdf
The Value of Certifying Products for FDO _ Paul at FIDO Alliance.pdfFIDO Alliance
 
IoT Analytics Company Presentation May 2024
IoT Analytics Company Presentation May 2024IoT Analytics Company Presentation May 2024
IoT Analytics Company Presentation May 2024IoTAnalytics
 
The Metaverse: Are We There Yet?
The  Metaverse:    Are   We  There  Yet?The  Metaverse:    Are   We  There  Yet?
The Metaverse: Are We There Yet?Mark Billinghurst
 
Structuring Teams and Portfolios for Success
Structuring Teams and Portfolios for SuccessStructuring Teams and Portfolios for Success
Structuring Teams and Portfolios for SuccessUXDXConf
 
10 Differences between Sales Cloud and CPQ, Blanka Doktorová
10 Differences between Sales Cloud and CPQ, Blanka Doktorová10 Differences between Sales Cloud and CPQ, Blanka Doktorová
10 Differences between Sales Cloud and CPQ, Blanka DoktorováCzechDreamin
 
UiPath Test Automation using UiPath Test Suite series, part 1
UiPath Test Automation using UiPath Test Suite series, part 1UiPath Test Automation using UiPath Test Suite series, part 1
UiPath Test Automation using UiPath Test Suite series, part 1DianaGray10
 
PLAI - Acceleration Program for Generative A.I. Startups
PLAI - Acceleration Program for Generative A.I. StartupsPLAI - Acceleration Program for Generative A.I. Startups
PLAI - Acceleration Program for Generative A.I. StartupsStefano
 
Powerful Start- the Key to Project Success, Barbara Laskowska
Powerful Start- the Key to Project Success, Barbara LaskowskaPowerful Start- the Key to Project Success, Barbara Laskowska
Powerful Start- the Key to Project Success, Barbara LaskowskaCzechDreamin
 
THE BEST IPTV in GERMANY for 2024: IPTVreel
THE BEST IPTV in  GERMANY for 2024: IPTVreelTHE BEST IPTV in  GERMANY for 2024: IPTVreel
THE BEST IPTV in GERMANY for 2024: IPTVreelreely ones
 
IESVE for Early Stage Design and Planning
IESVE for Early Stage Design and PlanningIESVE for Early Stage Design and Planning
IESVE for Early Stage Design and PlanningIES VE
 
Designing for Hardware Accessibility at Comcast
Designing for Hardware Accessibility at ComcastDesigning for Hardware Accessibility at Comcast
Designing for Hardware Accessibility at ComcastUXDXConf
 
Intro in Product Management - Коротко про професію продакт менеджера
Intro in Product Management - Коротко про професію продакт менеджераIntro in Product Management - Коротко про професію продакт менеджера
Intro in Product Management - Коротко про професію продакт менеджераMark Opanasiuk
 
Buy Epson EcoTank L3210 Colour Printer Online.pdf
Buy Epson EcoTank L3210 Colour Printer Online.pdfBuy Epson EcoTank L3210 Colour Printer Online.pdf
Buy Epson EcoTank L3210 Colour Printer Online.pdfEasyPrinterHelp
 
Introduction to FDO and How It works Applications _ Richard at FIDO Alliance.pdf
Introduction to FDO and How It works Applications _ Richard at FIDO Alliance.pdfIntroduction to FDO and How It works Applications _ Richard at FIDO Alliance.pdf
Introduction to FDO and How It works Applications _ Richard at FIDO Alliance.pdfFIDO Alliance
 
A Business-Centric Approach to Design System Strategy
A Business-Centric Approach to Design System StrategyA Business-Centric Approach to Design System Strategy
A Business-Centric Approach to Design System StrategyUXDXConf
 
Strategic AI Integration in Engineering Teams
Strategic AI Integration in Engineering TeamsStrategic AI Integration in Engineering Teams
Strategic AI Integration in Engineering TeamsUXDXConf
 
Agentic RAG What it is its types applications and implementation.pdf
Agentic RAG What it is its types applications and implementation.pdfAgentic RAG What it is its types applications and implementation.pdf
Agentic RAG What it is its types applications and implementation.pdfChristopherTHyatt
 
WSO2CONMay2024OpenSourceConferenceDebrief.pptx
WSO2CONMay2024OpenSourceConferenceDebrief.pptxWSO2CONMay2024OpenSourceConferenceDebrief.pptx
WSO2CONMay2024OpenSourceConferenceDebrief.pptxJennifer Lim
 
Connecting the Dots in Product Design at KAYAK
Connecting the Dots in Product Design at KAYAKConnecting the Dots in Product Design at KAYAK
Connecting the Dots in Product Design at KAYAKUXDXConf
 

Recently uploaded (20)

Simplified FDO Manufacturing Flow with TPMs _ Liam at Infineon.pdf
Simplified FDO Manufacturing Flow with TPMs _ Liam at Infineon.pdfSimplified FDO Manufacturing Flow with TPMs _ Liam at Infineon.pdf
Simplified FDO Manufacturing Flow with TPMs _ Liam at Infineon.pdf
 
The Value of Certifying Products for FDO _ Paul at FIDO Alliance.pdf
The Value of Certifying Products for FDO _ Paul at FIDO Alliance.pdfThe Value of Certifying Products for FDO _ Paul at FIDO Alliance.pdf
The Value of Certifying Products for FDO _ Paul at FIDO Alliance.pdf
 
IoT Analytics Company Presentation May 2024
IoT Analytics Company Presentation May 2024IoT Analytics Company Presentation May 2024
IoT Analytics Company Presentation May 2024
 
The Metaverse: Are We There Yet?
The  Metaverse:    Are   We  There  Yet?The  Metaverse:    Are   We  There  Yet?
The Metaverse: Are We There Yet?
 
Structuring Teams and Portfolios for Success
Structuring Teams and Portfolios for SuccessStructuring Teams and Portfolios for Success
Structuring Teams and Portfolios for Success
 
10 Differences between Sales Cloud and CPQ, Blanka Doktorová
10 Differences between Sales Cloud and CPQ, Blanka Doktorová10 Differences between Sales Cloud and CPQ, Blanka Doktorová
10 Differences between Sales Cloud and CPQ, Blanka Doktorová
 
UiPath Test Automation using UiPath Test Suite series, part 1
UiPath Test Automation using UiPath Test Suite series, part 1UiPath Test Automation using UiPath Test Suite series, part 1
UiPath Test Automation using UiPath Test Suite series, part 1
 
PLAI - Acceleration Program for Generative A.I. Startups
PLAI - Acceleration Program for Generative A.I. StartupsPLAI - Acceleration Program for Generative A.I. Startups
PLAI - Acceleration Program for Generative A.I. Startups
 
Powerful Start- the Key to Project Success, Barbara Laskowska
Powerful Start- the Key to Project Success, Barbara LaskowskaPowerful Start- the Key to Project Success, Barbara Laskowska
Powerful Start- the Key to Project Success, Barbara Laskowska
 
THE BEST IPTV in GERMANY for 2024: IPTVreel
THE BEST IPTV in  GERMANY for 2024: IPTVreelTHE BEST IPTV in  GERMANY for 2024: IPTVreel
THE BEST IPTV in GERMANY for 2024: IPTVreel
 
IESVE for Early Stage Design and Planning
IESVE for Early Stage Design and PlanningIESVE for Early Stage Design and Planning
IESVE for Early Stage Design and Planning
 
Designing for Hardware Accessibility at Comcast
Designing for Hardware Accessibility at ComcastDesigning for Hardware Accessibility at Comcast
Designing for Hardware Accessibility at Comcast
 
Intro in Product Management - Коротко про професію продакт менеджера
Intro in Product Management - Коротко про професію продакт менеджераIntro in Product Management - Коротко про професію продакт менеджера
Intro in Product Management - Коротко про професію продакт менеджера
 
Buy Epson EcoTank L3210 Colour Printer Online.pdf
Buy Epson EcoTank L3210 Colour Printer Online.pdfBuy Epson EcoTank L3210 Colour Printer Online.pdf
Buy Epson EcoTank L3210 Colour Printer Online.pdf
 
Introduction to FDO and How It works Applications _ Richard at FIDO Alliance.pdf
Introduction to FDO and How It works Applications _ Richard at FIDO Alliance.pdfIntroduction to FDO and How It works Applications _ Richard at FIDO Alliance.pdf
Introduction to FDO and How It works Applications _ Richard at FIDO Alliance.pdf
 
A Business-Centric Approach to Design System Strategy
A Business-Centric Approach to Design System StrategyA Business-Centric Approach to Design System Strategy
A Business-Centric Approach to Design System Strategy
 
Strategic AI Integration in Engineering Teams
Strategic AI Integration in Engineering TeamsStrategic AI Integration in Engineering Teams
Strategic AI Integration in Engineering Teams
 
Agentic RAG What it is its types applications and implementation.pdf
Agentic RAG What it is its types applications and implementation.pdfAgentic RAG What it is its types applications and implementation.pdf
Agentic RAG What it is its types applications and implementation.pdf
 
WSO2CONMay2024OpenSourceConferenceDebrief.pptx
WSO2CONMay2024OpenSourceConferenceDebrief.pptxWSO2CONMay2024OpenSourceConferenceDebrief.pptx
WSO2CONMay2024OpenSourceConferenceDebrief.pptx
 
Connecting the Dots in Product Design at KAYAK
Connecting the Dots in Product Design at KAYAKConnecting the Dots in Product Design at KAYAK
Connecting the Dots in Product Design at KAYAK
 

Composing graphical models with neural networks for structured representations and fast inference

  • 1. CS592 Presentation #18 Composing graphical models with neural networks for structured representations and fast inference 20173586 Jeongmin Cha 20184666 Yajie Zhou 20174463 Jaesung Choe
  • 2. Content 1. Motivation 2. Modeling idea 3. Structural Variational Autoencoder (SVAE) 4. Background 5. Main algorithm 6. Experiment 7. Group Discussion Point
  • 3. 1. Motivation ● How can we build interpretable models of high-dimensional data? ● modeling video of a mouse ● a mouse usually repeats a certain behavior dart groom rear
  • 4. 1. Motivation ● We want a model ○ can explain which behavior most is performing at each frame
  • 5. 1. Motivation ● What we want to do is ... ● segment and categorize mouse behavior from the video ● Q: generative vs discriminative model for this task?
  • 6. 1. Motivation ● What we want to do is ... ● segment and categorize mouse behavior from the video ● Q: generative vs discriminative model for this task? ○ We can use both ○ a large number of labeled data needed in discriminative scheme ○ discriminative model relaxes conditional independence assumption so may achieve better predictive result
  • 7. 1. Motivation ● What we want to do is ... ● segment and categorize mouse behavior from the video ● Q: generative vs discriminative model for this task? ○ We can use both ○ a large number of unlabeled data from a small number of labeled data
  • 8. 1. Motivation ● What we want to do is ... ● segment and categorize mouse behavior from the video ● Q: generative vs discriminative model for this task? ○ We can use both ○ a large number of unlabeled data from a small number of labeled data ○ This paper want to build a generative model for video of a mouse
  • 9. 1. Motivation ● a generative model for video of a mouse ● a mouse repeats certain behaviors GMM = one solution
  • 10. 1. Motivation ● However, what if the data is not well described by Gaussian?
  • 11. 1. Motivation ● a mixture of gaussians fits the data poorly ● reports too many clusters (not natural clustering result) GMMGMM
  • 12. 1. Motivation ● neural network fits data well ● but, difficult to interpret in high dimensions (lack interpretability) GMM density net (VAE)
  • 13. 1. Motivation ● neural network fits data well ● but, difficult to interpret in high dimensions (lack interpretability) ● does not explicitly represent discrete mixture components, GMM density net (VAE) An appropriate model might switch between discrete states
  • 14. 1. Motivation ● How about combining both? (Graphical model + Deep Learning) ● Structured Variational AutoEncoder (SVAE) density net (VAE) GMM GMM SVAE
  • 15. 1. Motivation ● Q: Graphical model vs Deep Learning, pros and cons?
  • 16. 1. Motivation ● Q: Graphical model vs Deep Learning, pros and cons? ● specify explicit relationship between variables before learning ○ Graphical model configuration starts from a higher level (deduction) ○ Deep learning configuration starts from a lower level (induction)
  • 17. 1. Motivation ● Graphical model ○ + interpretable, structured representations ○ + data and computational efficiency ○ - strong assumptions may not fit ○ - feature engineering ○ - top-down inference ● Deep learning ○ - not directly interpretable structure ○ - can require lots of data ○ + flexible representations, learn automatically ○ + feature learning ○ + recognition networks (bottom-up)
  • 18. 2. Modeling idea ● graphical models on latent variables ○ structured probability distributions ○ fast exact inference subroutines ● neural network models (VAE) for observations ○ produce flexible non-linear feature manifold ■ nonlinear high-dimensional data to low-dimensional and dense representations ○ recognition network ■ instead of learning variational distribution parameters directly ■ map observations to conjugate graphical model potentials ■
  • 19. 3. Structure Variational AutoEncoder (SVAE) Under the exponential conjugate property we can define SAVE as below, where p(θ) is the prior distribution and the p(x|θ) is the posterior distribution. Statistics function is defined and the partition function is described as, Finally, we should like to infer the likelihood.
  • 20. 3. Structure Variational AutoEncoder (SVAE) Under the exponential conjugate property we can define SAVE as below, where p(θ) is the prior distribution and the p(x|θ) is the posterior distribution. Statistics function is defined and the partition function is described as, Finally, we should like to infer the likelihood. Discussion Point: Can you tell the fundamental difference between VAE and SAVE ?
  • 21. 3. Structure Variational AutoEncoder (SVAE) Under the exponential conjugate property we can define SAVE as below, where p(θ) is the prior distribution and the p(x|θ) is the posterior distribution. Statistics function is defined and the partition function is described as, Finally, we should like to infer the likelihood. Discussion Point: Can you tell the fundamental difference between VAE and SAVE ? : Conjugate property
  • 22. 4. Background : conjugate distribution (VAE vs SVAE) if the posterior distributions p(x|θ) are in the same probability distribution family as the prior probability distribution p(θ), the prior and posterior are then called conjugate distributions. What is conjugate distribution? if the likelihood function is Poisson distribution, choosing a Poisson prior over the parameter λ will ensure that the posterior distribution is also Poisson distribution. Example where λ = 4 where k = 4 and θ = 1 Posteriori : Prior:
  • 23. 4. Background : conjugate distribution. if the posterior distributions p(x|θ) are in the same probability distribution family as the prior probability distribution p(θ), the prior and posterior are then called conjugate distributions. What is conjugate distribution? if the likelihood function is Poisson distribution, choosing a Poisson prior over the parameter λ will ensure that the posterior distribution is also Poisson distribution. Example Likelihood: (assume i=1, …, 6) If we set k = 10 and θ = 0.5, Prior becomes posterior: Prior:
  • 24. 4. Background : conjugate distribution. if the posterior distributions p(x|θ) are in the same probability distribution family as the prior probability distribution p(θ), the prior and posterior are then called conjugate distributions. What is conjugate distribution? estimate the likelihood by updating the parameters of our prior  —  reflecting a new mean and confidence level
  • 25. 4. Background : conjugate distribution. if the posterior distributions p(x|θ) are in the same probability distribution family as the prior probability distribution p(θ), the prior and posterior are then called conjugate distributions. What is conjugate distribution? estimate the likelihood by updating the parameters of our prior  —  reflecting a new mean and confidence level Discussion Point: Why the upper property is important in SVAE ?
  • 26. 4. Background : conjugate distribution. if the posterior distributions p(θ | x) are in the same probability distribution family as the prior probability distribution p(θ), the prior and posterior are then called conjugate distributions. What is conjugate distribution? estimate the likelihood by updating the parameters of our prior  —  reflecting a new mean and confidence level Discussion Point: Why the upper property is important in SVAE ? A : Conjugacy property is useful in Bayesian inference !
  • 27. 4. Background : conjugate distribution. the integral of the marginal likelihood = is intractable. Intractability in VAE. Conjugacy in SVAE (Proposition B.4) = where the posterior p(θ|x) is in the same exponential family as p(θ) with the natural parameter = , and d are statistic function.
  • 28. 4. Background : conjugate distribution. the integral of the marginal likelihood = is intractable. Intractability in VAE. Conjugacy in SVAE (Proposition B.4) = where the posterior p(θ|x) is in the same exponential family as p(θ) with the natural parameter = , and d are statistic function. VAE handles a general non-conjugate observation models by introducing recognition network.
  • 29. 4. Background : conjugate distribution. the integral of the marginal likelihood = is intractable. Intractability in VAE. Conjugacy in SVAE (Proposition B.4) where the posterior p(θ|x) is in the same exponential family as p(θ) = , and are statistic function.
  • 30. 4. Background : conjugate distribution. the integral of the marginal likelihood = is intractable. Intractability in VAE. Conjugacy in SVAE (Proposition B.4) = where the posterior p(θ|x) is in the same exponential family as p(θ) with the natural parameter = , and d are statistic function.
  • 31. 4. Background : conjugate distribution. the integral of the marginal likelihood = is intractable. Intractability in VAE. Conjugacy in SVAE (Proposition B.4) = = = This relationship is useful in Bayesian inference under the conjugacy property.
  • 35. 5.2 Non-conjugate Inference Discussion Point: is still linear Gaussian? what is conjugate prior?
  • 36. 5.2 Non-conjugate Inference SVAEs: recognition networks output conjugate potentials, then apply fast graphical model inference
  • 38. 5.2 Non-conjugate Inference Discussion Point: How to optimize this bound?
  • 43. https://www.youtube.com/watch?v=9WSb-89UsEo&t=60s (This video is disabled to be watched on other sites except for Youtube) 7. Group discussion Group Discussion Point: VAE vs SVAE : which model can have better performance? (Is it strike or ball?) : Strike : None : Strike : Ball ፠ Supplementary material - For those who are not familiar with the baseball rules.
  • 44. https://www.youtube.com/watch?v=9WSb-89UsEo&t=60s (This video is disabled to be watched on other sites except for Youtube) 7. Group discussion Group Discussion Point: VAE vs SVAE : which model can have better performance? (Is it strike or ball?) : Strike : None : Strike : Ball ፠ Supplementary material - For those who are not familiar with the baseball rules. Hint or not
  • 45. If SVAE follows the linear-chain structure, (expect) SVAE can predict better accuracy in video classification. (expect) VAE would be better for the single image classification. 7. Group discussion
  • 46. If SVAE follows the linear-chain structure, (expect) SVAE can predict better accuracy in video classification. (expect) VAE would be better for the single image classification. 7. Group discussion NO
  • 47. By the way, what is the result? strike - ball 7. Group discussion
  • 48. By the way, what is the result? strike - ball 7. Group discussion scoreboard
  • 49. By the way, what is the result? Strick!! 7. Group discussion strike - ball scoreboard
  • 50. By the way, what is the result? Strick!! How did you check the results? 7. Group discussion strike - ball scoreboard
  • 51. By the way, what is the result? Strick!! How did you check the results? I think just single frame is enough !! 7. Group discussion strike - ball scoreboard
  • 52. As we check the scoreboard, AI also look at the scoreboard for the inference. In other words, we do not need sequential frames. (Our expectation) VAE would be better. If we mask the scoreboard, (our expectation) SVAE would be better. 7. Group discussion Where AI is looking at (i.e. high attention) Mask(non-observable area)