arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2210.03728cs.LGcs.AIstat.ML
收藏

深度学习的动态潜空间分离

Dynamic Latent Separation for Deep Learning

University of California Santa Barbara(加州大学圣巴巴拉分校);University of California San Diego(加州大学圣迭戈分校)

Yi-Lin Tuan, Zih-Yun Chiu, William Yang Wang

AI总结:

提出动态潜空间分离方法atom modeling,通过在潜空间动态拉开样本距离增强输出多样性与部分可解释性,无需潜空间监督,有效提升了多任务中小到大规模模型的分类与生成性能。

摘要:

机器学习中的一个核心问题是,以灵活且可解释的方式学习表达性强的潜变量,用于对包含多个子组件的复杂数据进行模型预测。在此,我们开发了一种方法,它能提升表达性,提供部分解释,且不局限于特定应用。核心思想是在潜空间中动态拉开数据样本的距离,从而增强输出的多样性。受原子物理学启发,我们的动态潜空间分离方法依赖于每个数据样本联合学习的结构,这也揭示了每个子组件在区分数据样本时的重要性。这种方法被称为atom modeling,不需要对潜空间进行监督,并允许我们在模型原始目标之外学习额外的部分可解释表示。我们通过实验证明,该算法在多种分类和生成问题中也提升了小规模到较大规模模型的性能。

英文摘要:

A core problem in machine learning is to learn expressive latent variables for model prediction on complex data that involves multiple sub-components in a flexible and interpretable fashion. Here, we develop an approach that improves expressiveness, provides partial interpretation, and is not restricted to specific applications. The key idea is to dynamically distance data samples in the latent space and thus enhance the output diversity. Our dynamic latent separation method, inspired by atomic physics, relies on the jointly learned structures of each data sample, which also reveal the importance of each sub-component for distinguishing data samples. This approach, atom modeling, requires no supervision of the latent space and allows us to learn extra partially interpretable representations besides the original goal of a model. We empirically demonstrate that the algorithm also enhances the performance of small to larger-scale models in various classification and generation problems.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑