Discrete-space EOT/SB solving on High-dimensional Gaussian Mixture
95Conditional Shape ScoreDLightSB
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| DLightSBDimension (D)=2, Stochasticity parameter (gamma)=0.022025.09 | 95 | — | — | — | — | — | — | — | — | — | — | — | — | |
| DLightSBDimension (D)=2, Stochasticity parameter (gamma)=0.052025.09 | 92 | — | — | — | — | — | — | — | — | — | — | — | — | |
| CSBMDimension (D)=2, Stochasticity parameter (gamma)=0.02, Training Loss=KL, Step Count (N+1)=642025.09 | 88 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Feature-wise SBDimension (D)=2, Stochasticity parameter (gamma)=0.052025.09 | 87 | — | — | — | — | — | — | — | — | — | — | — | — | |
| CSBMDimension (D)=2, Stochasticity parameter (gamma)=0.05, Training Loss=KL, Step Count (N+1)=642025.09 | 87 | — | — | — | — | — | — | — | — | — | — | — | — | |
| α-CSBMDimension (D)=2, Stochasticity parameter (gamma)=0.05, Training Loss=KL, Step Count (N+1)=642025.09 | 87 | — | — | — | — | — | — | — | — | — | — | — | — | |
| IndependentDimension (D)=2, Stochasticity parameter (gamma)=0.052025.09 | 83 | — | — | — | — | — | — | — | — | — | — | — | — | |
| DLightSB-MDimension (D)=64, Stochasticity parameter (gamma)=0.02, Training Loss=KL, Step Count (N+1)=642025.09 | 83 | — | — | — | — | — | — | — | — | — | — | — | — | |
| CSBMDimension (D)=2, Stochasticity parameter (gamma)=0.05, Training Loss=MSE, Step Count (N+1)=642025.09 | 82 | — | — | — | — | — | — | — | — | — | — | — | — | |
| α-CSBMDimension (D)=2, Stochasticity parameter (gamma)=0.02, Training Loss=KL, Step Count (N+1)=642025.09 | 77 | — | — | — | — | — | — | — | — | — | — | — | — | |
| α-CSBMDimension (D)=2, Stochasticity parameter (gamma)=0.05, Training Loss=KL, Step Count (N+1)=162025.09 | 73 | — | — | — | — | — | — | — | — | — | — | — | — | |
| CSBMDimension (D)=2, Stochasticity parameter (gamma)=0.02, Training Loss=KL, Step Count (N+1)=162025.09 | 72 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Feature-wise SBDimension (D)=2, Stochasticity parameter (gamma)=0.022025.09 | 71 | — | — | — | — | — | — | — | — | — | — | — | — | |
| CSBMDimension (D)=2, Stochasticity parameter (gamma)=0.05, Training Loss=KL, Step Count (N+1)=162025.09 | 71 | — | — | — | — | — | — | — | — | — | — | — | — | |
| CSBMDimension (D)=2, Stochasticity parameter (gamma)=0.05, Training Loss=MSE, Step Count (N+1)=162025.09 | 71 | — | — | — | — | — | — | — | — | — | — | — | — | |
| α-CSBMDimension (D)=2, Stochasticity parameter (gamma)=0.02, Training Loss=KL, Step Count (N+1)=162025.09 | 66 | — | — | — | — | — | — | — | — | — | — | — | — | |
| IndependentDimension (D)=2, Stochasticity parameter (gamma)=0.022025.09 | 51 | — | — | — | — | — | — | — | — | — | — | — | — | |
| CSBMDimension (D)=2, Stochasticity parameter (gamma)=0.02, Training Loss=MSE, Step Count (N+1)=162025.09 | 48 | — | — | — | — | — | — | — | — | — | — | — | — | |
| ReferenceDimension (D)=2, Stochasticity parameter (gamma)=0.052025.09 | 45 | — | — | — | — | — | — | — | — | — | — | — | — | |
| CSBMDimension (D)=2, Stochasticity parameter (gamma)=0.02, Training Loss=MSE, Step Count (N+1)=642025.09 | 29 | — | — | — | — | — | — | — | — | — | — | — | — | |
| ReferenceDimension (D)=2, Stochasticity parameter (gamma)=0.022025.09 | 17 | — | — | — | — | — | — | — | — | — | — | — | — | |
| CSBMLoss=KL, N + 1=162025.09 | — | 77 | 72 | 92 | 91 | 82 | 78 | 85 | 79 | 91 | 90 | 84 | 89 | |
| CSBMLoss=KL, N + 1=642025.09 | — | 91 | 89 | 96 | 97 | 91 | 85 | 92 | 93 | 94 | 94 | 79 | 88 | |
| CSBMLoss=MSE, N + 1=162025.09 | — | 52 | 72 | 84 | 86 | 82 | 74 | 80 | 74 | 81 | 89 | 81 | 81 | |
| CSBMLoss=MSE, N + 1=642025.09 | — | 37 | 84 | 84 | 81 | 88 | 83 | 84 | 90 | 74 | 96 | 83 | 82 | |
| DLightSBLoss=-, N + 1=-2025.09 | — | 97 | 97 | 98 | 98 | 89 | 97 | 96 | 98 | 97 | 95 | 97 | 97 | |
| DLightSB-MLoss=KL, N + 1=162025.09 | — | 90 | 96 | 97 | 97 | 86 | 93 | 96 | 96 | 89 | 80 | 83 | 86 | |
| DLightSB-MLoss=KL, N + 1=642025.09 | — | 88 | 95 | 95 | 96 | 90 | 95 | 95 | 96 | 89 | 90 | 80 | 93 | |
| DLightSB-MLoss=MSE, N + 1=162025.09 | — | 75 | 95 | 85 | 93 | 69 | 96 | 93 | 92 | 72 | 88 | 89 | 79 | |
| DLightSB-MLoss=MSE, N + 1=642025.09 | — | 73 | 95 | 81 | 92 | 65 | 95 | 94 | 89 | 62 | 89 | 68 | 83 | |
| Feature-wise SBLoss=-, N + 1=-2025.09 | — | 92 | 99 | 98 | 98 | 92 | 98 | 98 | 98 | 98 | 98 | 98 | 98 | |
| IndependentLoss=-, N + 1=-2025.09 | — | 98 | 99 | 98 | 98 | 99 | 99 | 98 | 98 | 99 | 98 | 98 | 98 | |
| ReferenceLoss=-, N + 1=-2025.09 | — | 32 | 48 | 41 | 43 | 41 | 34 | 41 | 42 | 56 | 49 | 48 | 47 | |
| α-CSBMLoss=KL, N + 1=162025.09 | — | 74 | 75 | 93 | 91 | 79 | 81 | 85 | 78 | 93 | 89 | 81 | 87 | |
| α-CSBMLoss=KL, N + 1=642025.09 | — | 82 | 89 | 96 | 97 | 92 | 93 | 93 | 95 | 95 | 93 | 80 | 86 | |
| α-CSBMLoss=MSE, N + 1=162025.09 | — | 69 | 73 | 89 | 90 | 79 | 80 | 85 | 83 | 82 | 93 | 83 | 82 | |
| α-CSBMLoss=MSE, N + 1=642025.09 | — | 92 | 88 | 89 | 93 | 80 | 90 | 87 | 90 | 74 | 93 | 82 | 85 |