SlideShare ist ein Scribd-Unternehmen logo
1 von 20
Downloaden Sie, um offline zu lesen
A Wrapped Normal Distribution on Hyperbolic Space
for Gradient Based Learning
ICML’19, Jun 12th, 2019
Yoshihiro Nagano1), Shoichiro Yamaguchi2), Yasuhiro Fujita2), Masanori Koyama2)
1) Department of Complexity Science, The University of Tokyo, Japan
2) Preferred Networks, Inc., Japan
Paper: proceedings.mlr.press/v97/nagano19a.html
Code: github.com/pfnet-research/hyperbolic_wrapped_distribution
ICLR/ICML2019 , Jul 21st, 2019
Yoshihiro Nagano
2017-Current Ph.D. student @ UTokyo
Advisor: Masato Okada
Jul.-Sep. 2018 Summer Internship @ PFN
Mar. 2017 MSc. (Science) @ UTokyo
Mar. 2015 B.S. @ Keio Univ.
Interests
Generative Models, Neural Networks, Computational Neuroscience,
Unsupervised Learning
SNS
!: ganow.me / *: ganow / +: @ny_ganow
Motivation
ARTICLERESEARCH
Figure 3 | Monte Carlo tree search in AlphaGo. a, Each simulation
traverses the tree by selecting the edge with maximum action value Q,
plus a bonus u(P) that depends on a stored prior probability P for that
is evaluated
a rollout to
Selectiona b cExpansion Evaluation
p
p
Q + u(P)
Q + u(P)Q + u(P)
Q + u(P)
P P
P P
r
P
max
max
P
[Silver+2016]
Mammal
Primate
Human Monkey
Rodent
Motivation
Mammal
Primate
Human Monkey
Rodent
ARTICLECH
Monte Carlo tree search in AlphaGo. a, Each simulation
he tree by selecting the edge with maximum action value Q,
is evaluated in two ways: using the value network vθ
a rollout to the end of the game with the fast rollout
Selection b c dExpansion Evaluation Backup
p
p
Q + u(P)
Q + u(P)Q + u(P)
Q + u(P)
P P
P P
Q QQ
Q
rr r
P
max
max
P
[Silver+2016]
Hierarchical Datasets Hyperbolic Space
[Image: wikipedia.org]
[Nickel & Kiela, 2017]
Motivation
Mammal
Primate
Human Monkey
Rodent
ARTICLECH
Monte Carlo tree search in AlphaGo. a, Each simulation
he tree by selecting the edge with maximum action value Q,
is evaluated in two ways: using the value network vθ
a rollout to the end of the game with the fast rollout
Selection b c dExpansion Evaluation Backup
p
p
Q + u(P)
Q + u(P)Q + u(P)
Q + u(P)
P P
P P
Q QQ
Q
rr r
P
max
max
P
[Silver+2016]
Hierarchical Datasets Hyperbolic Space
Volume increases exponentially
with its radius
Motivation
Mammal
Primate
Human Monkey
Rodent
ARTICLECH
Monte Carlo tree search in AlphaGo. a, Each simulation
he tree by selecting the edge with maximum action value Q,
is evaluated in two ways: using the value network vθ
a rollout to the end of the game with the fast rollout
Selection b c dExpansion Evaluation Backup
p
p
Q + u(P)
Q + u(P)Q + u(P)
Q + u(P)
P P
P P
Q QQ
Q
rr r
P
max
max
P
[Silver+2016]
Hierarchical Datasets Hyperbolic Space
[Nickel+2017]
Motivation
Mammal
Primate
Human Monkey
Rodent
ARTICLECH
Monte Carlo tree search in AlphaGo. a, Each simulation
he tree by selecting the edge with maximum action value Q,
is evaluated in two ways: using the value network vθ
a rollout to the end of the game with the fast rollout
Selection b c dExpansion Evaluation Backup
p
p
Q + u(P)
Q + u(P)Q + u(P)
Q + u(P)
P P
P P
Q QQ
Q
rr r
P
max
max
P
[Silver+2016]
Hierarchical Datasets Hyperbolic Space
[Nickel+2017]
How can we extend these works to
probabilistic inference?
Difficulty: Probabilistic Distribution on Curved Space
…
M
1.
2. 3.
[Image: wikipedia.org]
Difficulty: Probabilistic Distribution on Curved Space
…
M
1.
2. 3.
[Image: wikipedia.org]
Difficulty: Probabilistic Distribution on Curved Space
…
M
1.
2. 3.
[Image: wikipedia.org]
Difficulty: Probabilistic Distribution on Curved Space
…
M
1.
2. 3.
[Image: wikipedia.org]
[ja.wikipedia.org]
(e.g. Poincaré disk, Lorentz model, …)
Lorentz Model
ℝ"#$ Lorentzian product
-1 n :
Hyperbolic Geometry
Hyperbolic Geometry
(Exponential Map)
(tangent space)
% ∈ '(ℍ* O
(Parallel Transport)
+ ∈ ',ℍ*
% ∈ '(ℍ*
Construction of Hyperbolic Wrapped Distribution
ℝ*
( )
Hyperbolic Wrapped Distribution(b)
Figure 3: The heatmaps of log-likelihood of the pesudo-
hyperbolic Gaussians with various µ and Σ. We designate
the origin of hyperbolic space by the × mark. See Ap-
pendix B for further details.
Since the metric at the tangent space coincides with the Eu-
clidean metric, we can produce various types of Hyperbolic
distributions by applying our construction strategy to other
distributions defined on Euclidean space, such as Laplace
and Cauchy distribution.
to a
rep
gra
wor
β-V
a sc
In H
is i
cod
µ
As
allo
dien
of t
rep
4.2
We
bili
lum
tual
wor
on
ing
wri
Density:
Projection:
(910 1
(;2 2 ; 9120 +
) 0 2 9 2 92 (
≃ ℝ* 2
Numerical Evaluations: VAEs on Synthetic Data
Hyperbolic VAE
Yoshihiro Nagano 1
Shoichiro Yamaguchi 2
Yasuhiro Fujita 2
Masanori Koyama 2
Abstract
rbolic space is a geometry that is known to
ell-suited for representation learning of data
an underlying hierarchical structure. In this
r, we present a novel hyperbolic distribution
d pseudo-hyperbolic Gaussian, a Gaussian-
distribution on hyperbolic space whose den-
can be evaluated analytically and differen-
d with respect to the parameters. Our dis-
ion enables the gradient-based learning of
robabilistic models on hyperbolic space that
d never have been considered before. Also,
an sample from this hyperbolic probability
bution without resorting to auxiliary means
ejection sampling. As applications of our
bution, we develop a hyperbolic-analog of
tional autoencoder and a method of prob-
tic word embedding on hyperbolic space.
emonstrate the efficacy of our distribution
rious datasets including MNIST, Atari 2600
kout, and WordNet.
duction
hyperbolic geometry is drawing attention as a
geometry to assist deep networks in capturing
tal structural properties of data such as a hi-
Hyperbolic attention network (G¨ulc¸ehre et al.,
proved the generalization performance of neural
on various tasks including machine translation
ng the hyperbolic geometry on several parts of
(a) A tree representation of the
training dataset
(b) Normal VAE (β = 1.0) (c) Hyperbolic VAE
Figure 1: The visual results of Hyperbolic VAE applied to
an artificial dataset generated by applying random pertur-
bations to a binary tree. The visualization is being done
on the Poincar´e ball. The red points are the embeddings
of the original tree, and the blue points are the embeddings
of noisy observations generated from the tree. The pink
× represents the origin of the hyperbolic space. The VAE
was trained without the prior knowledge of the tree struc-
ture. Please see 6.1 for experimental details
determines the properties of the dataset that can be learned
from the embedding. For the dataset with a hierarchical
stribution on Hyperbolic Space for
sed Learning
2
Yasuhiro Fujita 2
Masanori Koyama 2
(a) A tree representation of the
training dataset
(b) Normal VAE (β = 1.0) (c) Hyperbolic VAE
Figure 1: The visual results of Hyperbolic VAE applied to
(
Numerical Evaluations: VAEs on Breakout
Atari 2600 Breakout-v4
DQN [Mnih+ 2015]
VAE
(≒
)
Vanilla
Vanilla, |v|2 = 200
VanillaHyperbolic
Numerical Evaluations: Word Embeddings
WordNet Nouns word embedding
Euclid [Vilnis & McCallum 2015]
Conclusion
projection-based
hyperbolic wrapped distribution
VAE MNIST, Atari 2600
Breakout, WordNet
*: pfnet-research/hyperbolic_wrapped_distribution
+
Acknowledgements
Masaki
Watanabe
Tomohiro Hayase Kenta Oono
Takeru Miyato Sosuke Kobayashi
PFN2018

Weitere ähnliche Inhalte

Was ist angesagt?

NIPS読み会2013: One-shot learning by inverting a compositional causal process
NIPS読み会2013: One-shot learning by inverting  a compositional causal processNIPS読み会2013: One-shot learning by inverting  a compositional causal process
NIPS読み会2013: One-shot learning by inverting a compositional causal process
nozyh
 
Algorithms and tools for point cloud generation
Algorithms and tools for point cloud generationAlgorithms and tools for point cloud generation
Algorithms and tools for point cloud generation
Radhe Syam
 
REU-Airborn Toxins paper
REU-Airborn Toxins paperREU-Airborn Toxins paper
REU-Airborn Toxins paper
Sihan Chen
 
Efficient Data Stream Classification via Probabilistic Adaptive Windows
Efficient Data Stream Classification via Probabilistic Adaptive WindowsEfficient Data Stream Classification via Probabilistic Adaptive Windows
Efficient Data Stream Classification via Probabilistic Adaptive Windows
Albert Bifet
 

Was ist angesagt? (19)

Matching networks for one shot learning
Matching networks for one shot learningMatching networks for one shot learning
Matching networks for one shot learning
 
NIPS読み会2013: One-shot learning by inverting a compositional causal process
NIPS読み会2013: One-shot learning by inverting  a compositional causal processNIPS読み会2013: One-shot learning by inverting  a compositional causal process
NIPS読み会2013: One-shot learning by inverting a compositional causal process
 
Report
ReportReport
Report
 
Algorithms and tools for point cloud generation
Algorithms and tools for point cloud generationAlgorithms and tools for point cloud generation
Algorithms and tools for point cloud generation
 
A Hough Transform Based On a Map-Reduce Algorithm
A Hough Transform Based On a Map-Reduce AlgorithmA Hough Transform Based On a Map-Reduce Algorithm
A Hough Transform Based On a Map-Reduce Algorithm
 
safe and efficient off policy reinforcement learning
safe and efficient off policy reinforcement learningsafe and efficient off policy reinforcement learning
safe and efficient off policy reinforcement learning
 
REU-Airborn Toxins paper
REU-Airborn Toxins paperREU-Airborn Toxins paper
REU-Airborn Toxins paper
 
11 clusadvanced
11 clusadvanced11 clusadvanced
11 clusadvanced
 
The Gaussian Process Latent Variable Model (GPLVM)
The Gaussian Process Latent Variable Model (GPLVM)The Gaussian Process Latent Variable Model (GPLVM)
The Gaussian Process Latent Variable Model (GPLVM)
 
Interaction Networks for Learning about Objects, Relations and Physics
Interaction Networks for Learning about Objects, Relations and PhysicsInteraction Networks for Learning about Objects, Relations and Physics
Interaction Networks for Learning about Objects, Relations and Physics
 
[Paper] GIRAFFE: Representing Scenes as Compositional Generative Neural Featu...
[Paper] GIRAFFE: Representing Scenes as Compositional Generative Neural Featu...[Paper] GIRAFFE: Representing Scenes as Compositional Generative Neural Featu...
[Paper] GIRAFFE: Representing Scenes as Compositional Generative Neural Featu...
 
Efficient Data Stream Classification via Probabilistic Adaptive Windows
Efficient Data Stream Classification via Probabilistic Adaptive WindowsEfficient Data Stream Classification via Probabilistic Adaptive Windows
Efficient Data Stream Classification via Probabilistic Adaptive Windows
 
Dual Learning for Machine Translation (NIPS 2016)
Dual Learning for Machine Translation (NIPS 2016)Dual Learning for Machine Translation (NIPS 2016)
Dual Learning for Machine Translation (NIPS 2016)
 
Effective Numerical Computation in NumPy and SciPy
Effective Numerical Computation in NumPy and SciPyEffective Numerical Computation in NumPy and SciPy
Effective Numerical Computation in NumPy and SciPy
 
50120140502003
5012014050200350120140502003
50120140502003
 
Introduction to NumPy (PyData SV 2013)
Introduction to NumPy (PyData SV 2013)Introduction to NumPy (PyData SV 2013)
Introduction to NumPy (PyData SV 2013)
 
R and Visualization: A match made in Heaven
R and Visualization: A match made in HeavenR and Visualization: A match made in Heaven
R and Visualization: A match made in Heaven
 
Image Retrieval with Fisher Vectors of Binary Features (MIRU'14)
Image Retrieval with Fisher Vectors of Binary Features (MIRU'14)Image Retrieval with Fisher Vectors of Binary Features (MIRU'14)
Image Retrieval with Fisher Vectors of Binary Features (MIRU'14)
 
CLUSTERING HYPERSPECTRAL DATA
CLUSTERING HYPERSPECTRAL DATACLUSTERING HYPERSPECTRAL DATA
CLUSTERING HYPERSPECTRAL DATA
 

Ähnlich wie [ICLR/ICML2019読み会] A Wrapped Normal Distribution on Hyperbolic Space for Gradient Based Learning (ICML2019)

BALANCING BOARD MACHINES
BALANCING BOARD MACHINESBALANCING BOARD MACHINES
BALANCING BOARD MACHINES
butest
 
Deep Local Parametric Filters for Image Enhancement
Deep Local Parametric Filters for Image EnhancementDeep Local Parametric Filters for Image Enhancement
Deep Local Parametric Filters for Image Enhancement
Sean Moran
 

Ähnlich wie [ICLR/ICML2019読み会] A Wrapped Normal Distribution on Hyperbolic Space for Gradient Based Learning (ICML2019) (20)

Fast Object Recognition from 3D Depth Data with Extreme Learning Machine
Fast Object Recognition from 3D Depth Data with Extreme Learning MachineFast Object Recognition from 3D Depth Data with Extreme Learning Machine
Fast Object Recognition from 3D Depth Data with Extreme Learning Machine
 
Reversible Data Hiding in the Spatial and Frequency Domains
Reversible Data Hiding in the Spatial and Frequency DomainsReversible Data Hiding in the Spatial and Frequency Domains
Reversible Data Hiding in the Spatial and Frequency Domains
 
Gravitational Billion Body Project
Gravitational Billion Body ProjectGravitational Billion Body Project
Gravitational Billion Body Project
 
40120130406008
4012013040600840120130406008
40120130406008
 
Distance oracle - Truy vấn nhanh khoảng cách giữa hai điểm bất kỳ trên đồ thị
Distance oracle - Truy vấn nhanh khoảng cách giữa hai điểm bất kỳ trên đồ thịDistance oracle - Truy vấn nhanh khoảng cách giữa hai điểm bất kỳ trên đồ thị
Distance oracle - Truy vấn nhanh khoảng cách giữa hai điểm bất kỳ trên đồ thị
 
SciVisHalosFinalPaper
SciVisHalosFinalPaperSciVisHalosFinalPaper
SciVisHalosFinalPaper
 
BALANCING BOARD MACHINES
BALANCING BOARD MACHINESBALANCING BOARD MACHINES
BALANCING BOARD MACHINES
 
An introduction to deep learning
An introduction to deep learningAn introduction to deep learning
An introduction to deep learning
 
International Journal of Engineering and Science Invention (IJESI)
International Journal of Engineering and Science Invention (IJESI)International Journal of Engineering and Science Invention (IJESI)
International Journal of Engineering and Science Invention (IJESI)
 
T. Lucas Makinen x Imperial SBI Workshop
T. Lucas Makinen x Imperial SBI WorkshopT. Lucas Makinen x Imperial SBI Workshop
T. Lucas Makinen x Imperial SBI Workshop
 
Oc2423022305
Oc2423022305Oc2423022305
Oc2423022305
 
Estimation, Detection & Comparison of Soil Nutrients using Matlab
Estimation, Detection & Comparison of Soil Nutrients using MatlabEstimation, Detection & Comparison of Soil Nutrients using Matlab
Estimation, Detection & Comparison of Soil Nutrients using Matlab
 
Intepretability / Explainable AI for Deep Neural Networks
Intepretability / Explainable AI for Deep Neural NetworksIntepretability / Explainable AI for Deep Neural Networks
Intepretability / Explainable AI for Deep Neural Networks
 
Deep Local Parametric Filters for Image Enhancement
Deep Local Parametric Filters for Image EnhancementDeep Local Parametric Filters for Image Enhancement
Deep Local Parametric Filters for Image Enhancement
 
Graph Neural Network in practice
Graph Neural Network in practiceGraph Neural Network in practice
Graph Neural Network in practice
 
Introduction To Machine Learning and Neural Networks
Introduction To Machine Learning and Neural NetworksIntroduction To Machine Learning and Neural Networks
Introduction To Machine Learning and Neural Networks
 
www.ijerd.com
www.ijerd.comwww.ijerd.com
www.ijerd.com
 
Random Chaotic Number Generation based Clustered Image Encryption
Random Chaotic Number Generation based Clustered Image EncryptionRandom Chaotic Number Generation based Clustered Image Encryption
Random Chaotic Number Generation based Clustered Image Encryption
 
Classification of Iris Data using Kernel Radial Basis Probabilistic Neural N...
Classification of Iris Data using Kernel Radial Basis Probabilistic  Neural N...Classification of Iris Data using Kernel Radial Basis Probabilistic  Neural N...
Classification of Iris Data using Kernel Radial Basis Probabilistic Neural N...
 
Classification of Iris Data using Kernel Radial Basis Probabilistic Neural Ne...
Classification of Iris Data using Kernel Radial Basis Probabilistic Neural Ne...Classification of Iris Data using Kernel Radial Basis Probabilistic Neural Ne...
Classification of Iris Data using Kernel Radial Basis Probabilistic Neural Ne...
 

Kürzlich hochgeladen

development of diagnostic enzyme assay to detect leuser virus
development of diagnostic enzyme assay to detect leuser virusdevelopment of diagnostic enzyme assay to detect leuser virus
development of diagnostic enzyme assay to detect leuser virus
NazaninKarimi6
 
Human genetics..........................pptx
Human genetics..........................pptxHuman genetics..........................pptx
Human genetics..........................pptx
Silpa
 
Phenolics: types, biosynthesis and functions.
Phenolics: types, biosynthesis and functions.Phenolics: types, biosynthesis and functions.
Phenolics: types, biosynthesis and functions.
Silpa
 
The Mariana Trench remarkable geological features on Earth.pptx
The Mariana Trench remarkable geological features on Earth.pptxThe Mariana Trench remarkable geological features on Earth.pptx
The Mariana Trench remarkable geological features on Earth.pptx
seri bangash
 
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
?#DUbAI#??##{{(☎️+971_581248768%)**%*]'#abortion pills for sale in dubai@
 
Digital Dentistry.Digital Dentistryvv.pptx
Digital Dentistry.Digital Dentistryvv.pptxDigital Dentistry.Digital Dentistryvv.pptx
Digital Dentistry.Digital Dentistryvv.pptx
MohamedFarag457087
 
THE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptx
THE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptxTHE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptx
THE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptx
ANSARKHAN96
 
CYTOGENETIC MAP................ ppt.pptx
CYTOGENETIC MAP................ ppt.pptxCYTOGENETIC MAP................ ppt.pptx
CYTOGENETIC MAP................ ppt.pptx
Silpa
 

Kürzlich hochgeladen (20)

development of diagnostic enzyme assay to detect leuser virus
development of diagnostic enzyme assay to detect leuser virusdevelopment of diagnostic enzyme assay to detect leuser virus
development of diagnostic enzyme assay to detect leuser virus
 
Site Acceptance Test .
Site Acceptance Test                    .Site Acceptance Test                    .
Site Acceptance Test .
 
Selaginella: features, morphology ,anatomy and reproduction.
Selaginella: features, morphology ,anatomy and reproduction.Selaginella: features, morphology ,anatomy and reproduction.
Selaginella: features, morphology ,anatomy and reproduction.
 
Grade 7 - Lesson 1 - Microscope and Its Functions
Grade 7 - Lesson 1 - Microscope and Its FunctionsGrade 7 - Lesson 1 - Microscope and Its Functions
Grade 7 - Lesson 1 - Microscope and Its Functions
 
Human genetics..........................pptx
Human genetics..........................pptxHuman genetics..........................pptx
Human genetics..........................pptx
 
Phenolics: types, biosynthesis and functions.
Phenolics: types, biosynthesis and functions.Phenolics: types, biosynthesis and functions.
Phenolics: types, biosynthesis and functions.
 
FAIRSpectra - Enabling the FAIRification of Spectroscopy and Spectrometry
FAIRSpectra - Enabling the FAIRification of Spectroscopy and SpectrometryFAIRSpectra - Enabling the FAIRification of Spectroscopy and Spectrometry
FAIRSpectra - Enabling the FAIRification of Spectroscopy and Spectrometry
 
FAIRSpectra - Enabling the FAIRification of Analytical Science
FAIRSpectra - Enabling the FAIRification of Analytical ScienceFAIRSpectra - Enabling the FAIRification of Analytical Science
FAIRSpectra - Enabling the FAIRification of Analytical Science
 
Molecular markers- RFLP, RAPD, AFLP, SNP etc.
Molecular markers- RFLP, RAPD, AFLP, SNP etc.Molecular markers- RFLP, RAPD, AFLP, SNP etc.
Molecular markers- RFLP, RAPD, AFLP, SNP etc.
 
module for grade 9 for distance learning
module for grade 9 for distance learningmodule for grade 9 for distance learning
module for grade 9 for distance learning
 
The Mariana Trench remarkable geological features on Earth.pptx
The Mariana Trench remarkable geological features on Earth.pptxThe Mariana Trench remarkable geological features on Earth.pptx
The Mariana Trench remarkable geological features on Earth.pptx
 
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
 
Digital Dentistry.Digital Dentistryvv.pptx
Digital Dentistry.Digital Dentistryvv.pptxDigital Dentistry.Digital Dentistryvv.pptx
Digital Dentistry.Digital Dentistryvv.pptx
 
Atp synthase , Atp synthase complex 1 to 4.
Atp synthase , Atp synthase complex 1 to 4.Atp synthase , Atp synthase complex 1 to 4.
Atp synthase , Atp synthase complex 1 to 4.
 
Call Girls Ahmedabad +917728919243 call me Independent Escort Service
Call Girls Ahmedabad +917728919243 call me Independent Escort ServiceCall Girls Ahmedabad +917728919243 call me Independent Escort Service
Call Girls Ahmedabad +917728919243 call me Independent Escort Service
 
Role of AI in seed science Predictive modelling and Beyond.pptx
Role of AI in seed science  Predictive modelling and  Beyond.pptxRole of AI in seed science  Predictive modelling and  Beyond.pptx
Role of AI in seed science Predictive modelling and Beyond.pptx
 
GBSN - Biochemistry (Unit 2) Basic concept of organic chemistry
GBSN - Biochemistry (Unit 2) Basic concept of organic chemistry GBSN - Biochemistry (Unit 2) Basic concept of organic chemistry
GBSN - Biochemistry (Unit 2) Basic concept of organic chemistry
 
THE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptx
THE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptxTHE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptx
THE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptx
 
300003-World Science Day For Peace And Development.pptx
300003-World Science Day For Peace And Development.pptx300003-World Science Day For Peace And Development.pptx
300003-World Science Day For Peace And Development.pptx
 
CYTOGENETIC MAP................ ppt.pptx
CYTOGENETIC MAP................ ppt.pptxCYTOGENETIC MAP................ ppt.pptx
CYTOGENETIC MAP................ ppt.pptx
 

[ICLR/ICML2019読み会] A Wrapped Normal Distribution on Hyperbolic Space for Gradient Based Learning (ICML2019)

  • 1. A Wrapped Normal Distribution on Hyperbolic Space for Gradient Based Learning ICML’19, Jun 12th, 2019 Yoshihiro Nagano1), Shoichiro Yamaguchi2), Yasuhiro Fujita2), Masanori Koyama2) 1) Department of Complexity Science, The University of Tokyo, Japan 2) Preferred Networks, Inc., Japan Paper: proceedings.mlr.press/v97/nagano19a.html Code: github.com/pfnet-research/hyperbolic_wrapped_distribution ICLR/ICML2019 , Jul 21st, 2019
  • 2. Yoshihiro Nagano 2017-Current Ph.D. student @ UTokyo Advisor: Masato Okada Jul.-Sep. 2018 Summer Internship @ PFN Mar. 2017 MSc. (Science) @ UTokyo Mar. 2015 B.S. @ Keio Univ. Interests Generative Models, Neural Networks, Computational Neuroscience, Unsupervised Learning SNS !: ganow.me / *: ganow / +: @ny_ganow
  • 3. Motivation ARTICLERESEARCH Figure 3 | Monte Carlo tree search in AlphaGo. a, Each simulation traverses the tree by selecting the edge with maximum action value Q, plus a bonus u(P) that depends on a stored prior probability P for that is evaluated a rollout to Selectiona b cExpansion Evaluation p p Q + u(P) Q + u(P)Q + u(P) Q + u(P) P P P P r P max max P [Silver+2016] Mammal Primate Human Monkey Rodent
  • 4. Motivation Mammal Primate Human Monkey Rodent ARTICLECH Monte Carlo tree search in AlphaGo. a, Each simulation he tree by selecting the edge with maximum action value Q, is evaluated in two ways: using the value network vθ a rollout to the end of the game with the fast rollout Selection b c dExpansion Evaluation Backup p p Q + u(P) Q + u(P)Q + u(P) Q + u(P) P P P P Q QQ Q rr r P max max P [Silver+2016] Hierarchical Datasets Hyperbolic Space [Image: wikipedia.org] [Nickel & Kiela, 2017]
  • 5. Motivation Mammal Primate Human Monkey Rodent ARTICLECH Monte Carlo tree search in AlphaGo. a, Each simulation he tree by selecting the edge with maximum action value Q, is evaluated in two ways: using the value network vθ a rollout to the end of the game with the fast rollout Selection b c dExpansion Evaluation Backup p p Q + u(P) Q + u(P)Q + u(P) Q + u(P) P P P P Q QQ Q rr r P max max P [Silver+2016] Hierarchical Datasets Hyperbolic Space Volume increases exponentially with its radius
  • 6. Motivation Mammal Primate Human Monkey Rodent ARTICLECH Monte Carlo tree search in AlphaGo. a, Each simulation he tree by selecting the edge with maximum action value Q, is evaluated in two ways: using the value network vθ a rollout to the end of the game with the fast rollout Selection b c dExpansion Evaluation Backup p p Q + u(P) Q + u(P)Q + u(P) Q + u(P) P P P P Q QQ Q rr r P max max P [Silver+2016] Hierarchical Datasets Hyperbolic Space [Nickel+2017]
  • 7. Motivation Mammal Primate Human Monkey Rodent ARTICLECH Monte Carlo tree search in AlphaGo. a, Each simulation he tree by selecting the edge with maximum action value Q, is evaluated in two ways: using the value network vθ a rollout to the end of the game with the fast rollout Selection b c dExpansion Evaluation Backup p p Q + u(P) Q + u(P)Q + u(P) Q + u(P) P P P P Q QQ Q rr r P max max P [Silver+2016] Hierarchical Datasets Hyperbolic Space [Nickel+2017] How can we extend these works to probabilistic inference?
  • 8. Difficulty: Probabilistic Distribution on Curved Space … M 1. 2. 3. [Image: wikipedia.org]
  • 9. Difficulty: Probabilistic Distribution on Curved Space … M 1. 2. 3. [Image: wikipedia.org]
  • 10. Difficulty: Probabilistic Distribution on Curved Space … M 1. 2. 3. [Image: wikipedia.org]
  • 11. Difficulty: Probabilistic Distribution on Curved Space … M 1. 2. 3. [Image: wikipedia.org]
  • 12. [ja.wikipedia.org] (e.g. Poincaré disk, Lorentz model, …) Lorentz Model ℝ"#$ Lorentzian product -1 n : Hyperbolic Geometry
  • 13. Hyperbolic Geometry (Exponential Map) (tangent space) % ∈ '(ℍ* O (Parallel Transport) + ∈ ',ℍ* % ∈ '(ℍ*
  • 14. Construction of Hyperbolic Wrapped Distribution ℝ* ( )
  • 15. Hyperbolic Wrapped Distribution(b) Figure 3: The heatmaps of log-likelihood of the pesudo- hyperbolic Gaussians with various µ and Σ. We designate the origin of hyperbolic space by the × mark. See Ap- pendix B for further details. Since the metric at the tangent space coincides with the Eu- clidean metric, we can produce various types of Hyperbolic distributions by applying our construction strategy to other distributions defined on Euclidean space, such as Laplace and Cauchy distribution. to a rep gra wor β-V a sc In H is i cod µ As allo dien of t rep 4.2 We bili lum tual wor on ing wri Density: Projection: (910 1 (;2 2 ; 9120 + ) 0 2 9 2 92 ( ≃ ℝ* 2
  • 16. Numerical Evaluations: VAEs on Synthetic Data Hyperbolic VAE Yoshihiro Nagano 1 Shoichiro Yamaguchi 2 Yasuhiro Fujita 2 Masanori Koyama 2 Abstract rbolic space is a geometry that is known to ell-suited for representation learning of data an underlying hierarchical structure. In this r, we present a novel hyperbolic distribution d pseudo-hyperbolic Gaussian, a Gaussian- distribution on hyperbolic space whose den- can be evaluated analytically and differen- d with respect to the parameters. Our dis- ion enables the gradient-based learning of robabilistic models on hyperbolic space that d never have been considered before. Also, an sample from this hyperbolic probability bution without resorting to auxiliary means ejection sampling. As applications of our bution, we develop a hyperbolic-analog of tional autoencoder and a method of prob- tic word embedding on hyperbolic space. emonstrate the efficacy of our distribution rious datasets including MNIST, Atari 2600 kout, and WordNet. duction hyperbolic geometry is drawing attention as a geometry to assist deep networks in capturing tal structural properties of data such as a hi- Hyperbolic attention network (G¨ulc¸ehre et al., proved the generalization performance of neural on various tasks including machine translation ng the hyperbolic geometry on several parts of (a) A tree representation of the training dataset (b) Normal VAE (β = 1.0) (c) Hyperbolic VAE Figure 1: The visual results of Hyperbolic VAE applied to an artificial dataset generated by applying random pertur- bations to a binary tree. The visualization is being done on the Poincar´e ball. The red points are the embeddings of the original tree, and the blue points are the embeddings of noisy observations generated from the tree. The pink × represents the origin of the hyperbolic space. The VAE was trained without the prior knowledge of the tree struc- ture. Please see 6.1 for experimental details determines the properties of the dataset that can be learned from the embedding. For the dataset with a hierarchical stribution on Hyperbolic Space for sed Learning 2 Yasuhiro Fujita 2 Masanori Koyama 2 (a) A tree representation of the training dataset (b) Normal VAE (β = 1.0) (c) Hyperbolic VAE Figure 1: The visual results of Hyperbolic VAE applied to (
  • 17. Numerical Evaluations: VAEs on Breakout Atari 2600 Breakout-v4 DQN [Mnih+ 2015] VAE (≒ ) Vanilla Vanilla, |v|2 = 200 VanillaHyperbolic
  • 18. Numerical Evaluations: Word Embeddings WordNet Nouns word embedding Euclid [Vilnis & McCallum 2015]
  • 19. Conclusion projection-based hyperbolic wrapped distribution VAE MNIST, Atari 2600 Breakout, WordNet *: pfnet-research/hyperbolic_wrapped_distribution +
  • 20. Acknowledgements Masaki Watanabe Tomohiro Hayase Kenta Oono Takeru Miyato Sosuke Kobayashi PFN2018