at the boundary between light transport simulation …hachisuka/mcqmc2014.pdflight transport...
TRANSCRIPT
![Page 1: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/1.jpg)
At the Boundary Between Light Transport Simulation and
Computational Statistics
Toshiya Hachisuka
Aarhus University
MCQMC2014
![Page 2: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/2.jpg)
At the boundary
2
Ij = E
fj(X)
p(X)
Me
![Page 3: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/3.jpg)
What’s often considered
• Graphics communities “just use” statistics
3
StatisticsGraphics
![Page 4: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/4.jpg)
What I hope to see
• Statistics communities take something from graphics
4
StatisticsGraphics
![Page 5: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/5.jpg)
Aim of this talk
• Motivate you to facilitate cross-pollination
• Both ways (graphics statistics)
• Using two examples
• Progressive density estimation
• Primary space serial tempering
5
“Progressive Photon Mapping” Hachisuka et al. ACM SIGGRAPH Asia 2008
“Multiplexed Metropolis Light Transport” Hachisuka et al. ACM SIGGRAPH 2014 (conditionally accepted)
![Page 6: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/6.jpg)
6
Progressive Density Estimation
![Page 7: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/7.jpg)
7
Manifold of non-zero valuesΩ∞
L =
Ω∞
fjXdµ(X)
![Page 8: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/8.jpg)
L =
Ω∞
fjXdµ(X)
8
Ω∞
xi ∼ p(xi)≈ 1
N
i
fj (xi)
p(xi)
![Page 9: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/9.jpg)
9
Eye
Transparent Layer
Light source
Matte Surface
![Page 10: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/10.jpg)
10
![Page 11: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/11.jpg)
11
θ1
![Page 12: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/12.jpg)
12
θ1
![Page 13: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/13.jpg)
13
θ2
θ1
![Page 14: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/14.jpg)
14
θ2
θ1
![Page 15: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/15.jpg)
15
Probability of sampling this path is nearly zero
θ2
θ1
![Page 16: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/16.jpg)
16
Probability of sampling this path is nearly zero
θ2
θ1
![Page 17: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/17.jpg)
17
Probability of hitting at the same point is zero
θ2 θ1
![Page 18: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/18.jpg)
Manifold of non-samplable paths
18
θ1
θ2
![Page 19: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/19.jpg)
19
![Page 20: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/20.jpg)
20
![Page 21: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/21.jpg)
21
Light source
Eye
Glass
Matte surface
Matte surface
![Page 22: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/22.jpg)
22
How can we capture non-samplable paths?
![Page 23: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/23.jpg)
23
Progressive Density Estimation
![Page 24: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/24.jpg)
24
θ2 θ1
![Page 25: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/25.jpg)
25
θ2 θ1
Allow approximate connection within distance
r
r
![Page 26: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/26.jpg)
26
θ2 θ1
Allow approximate connection within distance r... and progressively reduce r
![Page 27: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/27.jpg)
Progressive density estimation
• Algorithmically convergent density estimation
• Add gradually reducing bias to relax the problem
• First to solve the issue of non-samplable paths
27
[Hachisuka et al. 2008]
![Page 28: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/28.jpg)
28
Initial pass
Image
![Page 29: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/29.jpg)
29
Initial pass
j-th pixel
![Page 30: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/30.jpg)
30
Initial pass
xj Visible point
![Page 31: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/31.jpg)
31
Initial pass
Kernelr0
r0Initial kernel bandwidth:
![Page 32: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/32.jpg)
32
1st passMonte Carlo photon tracing
![Page 33: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/33.jpg)
33
1st pass
Photons
![Page 34: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/34.jpg)
34
1st pass
Range query with r0
![Page 35: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/35.jpg)
35
1st pass
Reduce radius into r1
![Page 36: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/36.jpg)
Radius reduction
• Reduce radius based on sample statistics
• Number of photons found so far
• Number of newly found photons
• Reduction rate (user-defined)
36
rn+1 =Nn + αMn
Nn +Mnrn
Nn
Mn
α ∈]0, 1[
Nn+1 = Nn + αMn
![Page 37: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/37.jpg)
37
1st pass
r1 =N0 + αM0
N0 +M0r0 =
0 + α4
0 + 4r0
![Page 38: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/38.jpg)
38
1st pass
N1 = N0 + αM0 = 0 + α4
![Page 39: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/39.jpg)
39
2nd pass
![Page 40: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/40.jpg)
40
2nd pass
New set of photons
![Page 41: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/41.jpg)
41
2nd pass
Range query with r1
![Page 42: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/42.jpg)
42
2nd pass
r2 =N1 + αM1
N1 +M1r1 N2 = N1 + αM1
![Page 43: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/43.jpg)
43
3rd and n-th passes
rn+1 =Nn + αMn
Nn +Mnrn Nn+1 = Nn + αMn
![Page 44: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/44.jpg)
44
MC integration Progressive density estimation
Equal time
![Page 45: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/45.jpg)
Convergence
45
• Converges to the correct solution
• Infinite number of neighboring samples
• Infinitely small radius
L = limn→∞
Nn
πr2n
![Page 46: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/46.jpg)
• Converges to the correct solution
•
•
46
Convergence
limi→∞
Ni(Ri) = ∞limi→∞
Ri(Ni) = 0
L = limn→∞
Nn
πr2n
![Page 47: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/47.jpg)
• Converges to the correct solution
•
•
• Convergence rate
• Variance
• Bias
47
Convergence
limi→∞
Ni(Ri) = ∞limi→∞
Ri(Ni) = 0
L = limn→∞
Nn
πr2n
O(n−α)
O(nα−1)α =
2
3Balanced with
![Page 48: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/48.jpg)
Relationship to recursive density estimation
• Recursive density estimation
• Predefined sequence of radii
• Assume stationary density distribution
• Progressive density estimation
• Adaptive sequence from sample statistics
• No assumption on stationarity
48
[Wolverton and Wagner 1969, Yamato 1971, Knaus and Zwicker 2011]
[Hachisuka et al. 2008, Hachisuka and Jensen 2009]
![Page 49: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/49.jpg)
Progressive vs Ordinary
49
Progressive Ordinary
Computation Unbounded Unbounded
Storage Bounded Unbounded
Convergence Single run Multiple runs
![Page 50: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/50.jpg)
50
Primary Space Serial Tempering
![Page 51: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/51.jpg)
Multiple sampling methods
51
![Page 52: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/52.jpg)
Multiple sampling methods
52
![Page 53: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/53.jpg)
Multiple sampling methods
53
![Page 54: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/54.jpg)
Multiple sampling methods
54
![Page 55: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/55.jpg)
Multiple sampling methods
55
p0(X)
p1(X) p2(X)
p3(X)
Given the same integrand, there are many known PDFs
![Page 56: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/56.jpg)
Multiple importance sampling
• Utilizes samples from multiple PDFs by weighting
56
[Veach and Guibas 1995]
f(X)
![Page 57: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/57.jpg)
Multiple importance sampling
• Utilizes samples from multiple PDFs by weighting
57
[Veach and Guibas 1995]
p0(X) p1(X)
f(X)
![Page 58: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/58.jpg)
Multiple importance sampling
• Importance sampling
• Multiple importance sampling
58
[Veach and Guibas 1995]
f(X)dµ(X) ≈ 1
N
i
f(xi)
p0(xi)p0(xi)
f(X)dµ(X) ≈ 1
N
w0(xi)
f(xi)
p0(xi)+ w1(xi)
f(xi)
p1(xi)
p0(xi) p1(xi)
1
N
i
f(xi)
p1(xi)p1(xi)or
Uses both
![Page 59: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/59.jpg)
Markov chain Monte Carlo sampling
• Requires no knowledge of PDFs
59
f(X)
![Page 60: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/60.jpg)
60
How can we combine MIS and MCMC?
![Page 61: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/61.jpg)
Combination
• Multiple importance sampling
• Ordinary Monte Carlo method
• Utilizes the knowledge of all the PDFs
• Markov chain Monte Carlo sampling
• Markov chain Monte Carlo method
• No usage (or requirement) of PDFs
61
![Page 62: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/62.jpg)
62
Primary Space Serial Tempering
![Page 63: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/63.jpg)
Primary sample space
• Hypercube of random numbers
• Mapping from a point to a sample
• Inverse CDF = mapping function
63
[Kelemen et al. 2002]
U ∈ ]0, 1[N
C(U) =f(P−1(U))
p(P−1(U))
f(X)dµ(X) ≈ 1
N
i
f(xi)
p(xi)=
1
N
i
C(ui)
![Page 64: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/64.jpg)
Primary sample space
• Hypercube of random numbers
• Mapping from a point to a sample
• Inverse CDF = mapping function
64
[Kelemen et al. 2002]
Point in the hypercubePath in the sample space
U ∈ ]0, 1[N
X = P−1(U)X U
![Page 65: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/65.jpg)
Serial tempering
• MCMC sampling of a sum of multiple distributions
• Extended states of the Markov chain
• Index j is updated via MCMC
• Extended state is originally temperature
• Still does not use any PDFs
65
x ∼
j
fj(x)
(x, j)
[Marinari and Parisi 1992, Geyer and Thompson 1995]
![Page 66: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/66.jpg)
Primary space serial tempering
• Use a weighted sum of primary space distributions
• Extended states is now
• Index j is the index to the j-th PDF/CDF
66
u ∼
j
wj(u)Cj(u)
wj(U) = w(P−1j (U))
(u, j)
Cj(U) =f(P−1
j (U))
pj(P−1j (U))
[Hachisuka et al. 2014] (TBA)
MIS weightPrimary space distribution
![Page 67: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/67.jpg)
Original primary space
67
C(U)
x = P−1(U)
![Page 68: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/68.jpg)
Multiplexed primary space
68
w3(U)C3(U)w2(U)C2(U)w1(U)C1(U)w0(U)C0(U)
x = P−1j (U)
![Page 69: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/69.jpg)
Primary space serial tempering
69
w3(U)C3(U)w2(U)C2(U)w1(U)C1(U)w0(U)C0(U)
x = P−10 (U)
![Page 70: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/70.jpg)
Primary space serial tempering
70
w3(U)C3(U)w2(U)C2(U)w1(U)C1(U)w0(U)C0(U)
x = P−10 (U)
![Page 71: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/71.jpg)
Primary space serial tempering
71
w3(U)C3(U)w2(U)C2(U)w1(U)C1(U)w0(U)C0(U)
x = P−11 (U)
![Page 72: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/72.jpg)
Primary space serial tempering
72
w3(U)C3(U)w2(U)C2(U)w1(U)C1(U)w0(U)C0(U)
x = P−13 (U)
![Page 73: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/73.jpg)
Primary space serial tempering
73
w3(U)C3(U)w2(U)C2(U)w1(U)C1(U)w0(U)C0(U)
x = P−13 (U)
![Page 74: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/74.jpg)
[Veach and Guibas 1997]Metropolis-Hastings
74
Equal time
![Page 75: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/75.jpg)
Primary space only
75
[Kelemen et al. 2002]
Equal time
![Page 76: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/76.jpg)
Primary space serial tempering
76
Equal time
![Page 77: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/77.jpg)
[Veach and Guibas 1997]Metropolis-Hastings
77
Equal time
![Page 78: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/78.jpg)
78
Summary
![Page 79: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/79.jpg)
Statistics / Graphics
• Progressive density estimation
• Statistics: new density estimator
• Graphics: simulation of complex light paths
• Primary space serial tempering
• Statistics: MCMC driven MIS
• Graphics: robust and efficient path sampling
79
![Page 80: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/80.jpg)
What I hope to see
• Statistics communities take something from graphics
80
StatisticsGraphics
![Page 81: At the Boundary Between Light Transport Simulation …hachisuka/mcqmc2014.pdfLight Transport Simulation and Computational Statistics Toshiya Hachisuka Aarhus University MCQMC2014 At](https://reader033.vdocuments.site/reader033/viewer/2022060310/5f0a9acd7e708231d42c736b/html5/thumbnails/81.jpg)
StatisticsGraphics
What I really hope to see
• We collaborate to develop something interesting!
81