论文标题
深残余混合模型
Deep Residual Mixture Models
论文作者
论文摘要
我们提出了深层混合模型（DRMMS），这是一种新型的深层生成模型结构。与其他深层模型相比，DRMM允许更灵活的条件采样：该模型可以使用所有变量训练一次，然后用于与调节变量，高斯先验和（IN）等效约束的任意组合进行采样。这为交互式和探索性机器学习提供了新的机会，其中应将等待重新培训的用户最小化。我们演示了受约束的多LIMB逆运动学和可控生成动画中的DRMM。
We propose Deep Residual Mixture Models (DRMMs), a novel deep generative model architecture. Compared to other deep models, DRMMs allow more flexible conditional sampling: The model can be trained once with all variables, and then used for sampling with arbitrary combinations of conditioning variables, Gaussian priors, and (in)equality constraints. This provides new opportunities for interactive and exploratory machine learning, where one should minimize the user waiting for retraining a model. We demonstrate DRMMs in constrained multi-limb inverse kinematics and controllable generation of animations.
