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 2209.01258cs.AIq-bio.NC
收藏

基于对象的主动推理

Object-based active inference

Radboud University(拉德堡德大学);Donders Institute for Brain, Cognition & Behavior(唐德斯脑、认知与行为研究所)

Ruben S. van Bergen, Pablo L. Lanillos

摘要:

世界由物体组成:具有独立属性和动力的不同实体。为了让智能体与世界智能地交互,它们必须将感官输入转化为描述每个对象的绑定在一起的特征。这些基于对象的表示形成了规划行为的自然基础。主动推理(AIF)是一种有影响力的感知和行动的统一解释,但现有的AIF模型没有利用这一重要的归纳偏差。为了解决这个问题,我们引入了“基于对象的主动推理”(OBAI),将AIF与最近的基于深度对象的神经网络结合起来。OBAI表示不同的对象与单独的变分信念,并使用选择性注意路由输入到相应的对象插槽。对象表示被赋予独立的基于动作的动力学。动力学和生成模型是从简单环境(活动多dSprites)的经验中学习的。我们表明,OBAI学习正确分割的动作扰动对象从视频输入,并操纵这些对象向任意目标。

英文摘要:

The world consists of objects: distinct entities possessing independent properties and dynamics. For agents to interact with the world intelligently, they must translate sensory inputs into the bound-together features that describe each object. These object-based representations form a natural basis for planning behavior. Active inference (AIF) is an influential unifying account of perception and action, but existing AIF models have not leveraged this important inductive bias. To remedy this, we introduce 'object-based active inference' (OBAI), marrying AIF with recent deep object-based neural networks. OBAI represents distinct objects with separate variational beliefs, and uses selective attention to route inputs to their corresponding object slots. Object representations are endowed with independent action-based dynamics. The dynamics and generative model are learned from experience with a simple environment (active multi-dSprites). We show that OBAI learns to correctly segment the action-perturbed objects from video input, and to manipulate these objects towards arbitrary goals.

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。

↑