Diffusion models have achieved justifiable popularity by attaining
state-of-the-art performance in generating realistic objects, including when
conditioning generation on labels. Current diffusion models are universally
linear in nature, modeling diffusion identically for objects of all classes.
For the multi-class conditional generation problem, we propose a novel,
structurally unique framework of diffusion models which are hierarchically
branched according to the inherent relationships between classes. In this work,
we showcase several advantages of branched diffusion models. We demonstrate
that branched models generate samples more efficiently, and are more easily
extended to novel classes in a continual-learning setting. We also show that
branched models enjoy a unique interpretability that offers insight into the
modeled data distribution. Branched diffusion models represent an alternative
paradigm to their traditional linear counterparts, and can have large impacts
in how we use diffusion models for efficient generation, online learning, and
scientific discovery.

本文提出了一种基于类别之间内在关系建立层级分支的扩散模型框架，证明了相比于线性模型，分支模型在多类别条件生成问题中更加高效，并且更容易延伸到新类别中。同时，这些分支模型还具有独特的可解释性，可以对建模的数据分布提供深入洞见，展望分支扩散模型在高效生成、在线学习和学术发现方面的广泛用途。

层次分支扩散模型实现高效且可解释的多类别条件生成

Hierarchically branched diffusion models for efficient and interpretable multi-class conditional generation

We explore the behavior of a standard convolutional neural net in a
continual-learning setting that introduces visual classification tasks
sequentially and requires the net to master new tasks while preserving mastery
of previously learned tasks. This setting corresponds to that which human
learners face as they acquire domain expertise serially, for example, as an
individual studies a textbook. Through simulations involving sequences of ten
related visual tasks, we find reason for optimism that nets will scale well as
they advance from having a single skill to becoming multi-skill domain experts.
We observe two key phenomena. First, \emph{forward facilitation}---the
accelerated learning of task $n+1$ having learned $n$ previous tasks---grows
with $n$. Second, \emph{backward interference}---the forgetting of the $n$
previous tasks when learning task $n+1$---diminishes with $n$. Amplifying
forward facilitation is the goal of research on metalearning, and attenuating
backward interference is the goal of research on catastrophic forgetting. We
find that both of these goals are attained simply through broader exposure to a
domain.

我们研究了卷积神经网络在连续学习环境中的行为，通常是按顺序引入视觉分类任务并要求网络在掌握以前学习的任务的同时掌握新任务。我们发现，在涉及十个相关视觉任务的模拟中，前向促进随着 n 的增加而增加，而向后干涉则随 n 的增加而减少。这可以通过更广泛地接触某一领域来实现元学习的研究目标，并通过减少灾难性遗忘的研究来实现。