Statistical Learning
Amortized Projection Optimization for Sliced Wasserstein Generative Models
However, finding these directions usually requires an iterative optimization procedure over the space of projecting directions, which is computationally expensive. Moreover, the computational issue is even more severe in deep learning applications, where computing the distance between two mini-batch probability measures is repeated several times.
Supplementary Material for PDSketch: Integrated Planning Domain Programming and Learning Jiayuan Mao 1 Tomás Lozano-Pérez 1 Joshua B. T enenbaum 1,2,3 Leslie Pack Kaelbling 1 1
We have repeated all experiments three times.
Predicting Label Distribution from Multi-label Ranking
There are currently two strategies including LDL (label distribution learning) and LE (label enhancement) to predict label distributions.
Estimating the Arc Length of the Optimal ROC Curve and Lower Bounding the Maximal AUC Song Liu
In applications that require computing statistical discrepancy between distributions (e.g.
Supplementary Material for " Learning Superpoint Graph Cut for 3D Instance Segmentation " Le Hui
This supplementary material provides more details on network architecture, visualization, and ablation study of our method. We also analyze the limitation and discuss the impact of our method. D, we discuss the limitations and impacts of our method. The produced feature dimension of 3D U-Net is 32. Finally, we can obtain 32-dimensional super-point features.
