论文标题
自回归得分匹配
Autoregressive Score Matching
论文作者
论文摘要
自回归模型使用链条规则将关节概率分布定义为条件性的产物。这些条件需要归一化，对可以使用的功能家庭施加限制。为了提高灵活性，我们提出了自回归的条件分数模型（AR-CSM），在该模型中，我们以单变量对数 - 条件（分数）（得分）的衍生物为角度对关节分布进行参数化，这不必归一化。为了训练AR-CSM，我们在名为复合分数匹配（CSM）之间引入了新的分歧。对于AR-CSM模型，可以有效地计算和优化数据和模型分布之间的这种差异，不需要昂贵的采样或对抗性训练。与以前的分数匹配算法相比，我们的方法更可扩展到高维数据，并且更稳定以优化。我们通过广泛的实验结果显示，它可以应用于与隐式编码器的合成数据，图像产生，图像降解和训练潜在变量模型的密度估计。
Autoregressive models use chain rule to define a joint probability distribution as a product of conditionals. These conditionals need to be normalized, imposing constraints on the functional families that can be used. To increase flexibility, we propose autoregressive conditional score models (AR-CSM) where we parameterize the joint distribution in terms of the derivatives of univariate log-conditionals (scores), which need not be normalized. To train AR-CSM, we introduce a new divergence between distributions named Composite Score Matching (CSM). For AR-CSM models, this divergence between data and model distributions can be computed and optimized efficiently, requiring no expensive sampling or adversarial training. Compared to previous score matching algorithms, our method is more scalable to high dimensional data and more stable to optimize. We show with extensive experimental results that it can be applied to density estimation on synthetic data, image generation, image denoising, and training latent variable models with implicit encoders.
