arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1612.00341cs.AIcs.LG
收藏

基于组合对象的物理动力学学习方法

A Compositional Object-Based Approach to Learning Physical Dynamics

MIT(麻省理工学院)

Michael B. Chang, Tomer Ullman, Antonio Torralba, Joshua B. Tenenbaum

摘要:

我们提出了神经物理引擎(NPE),这是一个用于学习直观物理模拟器的框架,它可以自然地概括可变的对象数量和不同的场景配置。我们提出了一个可组合的基于对象的表示和一个神经网络架构,其组成结构分解成成对的相互作用的对象动态的一个物理场景的分解。像一个符号物理引擎一样,NPE被赋予了对象及其相互作用的通用概念;实现为神经网络,它可以通过随机梯度下降进行训练,以适应不同世界的特定对象属性和动态。我们评估我们的方法在二维世界中的简单刚体动力学的功效。通过比较结构化程度较低的架构,我们表明,NPE的组成表示的结构在物理相互作用提高了其预测运动的能力,概括在变量对象计数和不同的场景配置,并推断潜在的属性的对象,如质量。

英文摘要:

We present the Neural Physics Engine (NPE), a framework for learning simulators of intuitive physics that naturally generalize across variable object count and different scene configurations. We propose a factorization of a physical scene into composable object-based representations and a neural network architecture whose compositional structure factorizes object dynamics into pairwise interactions. Like a symbolic physics engine, the NPE is endowed with generic notions of objects and their interactions; realized as a neural network, it can be trained via stochastic gradient descent to adapt to specific object properties and dynamics of different worlds. We evaluate the efficacy of our approach on simple rigid body dynamics in two-dimensional worlds. By comparing to less structured architectures, we show that the NPE's compositional representation of the structure in physical interactions improves its ability to predict movement, generalize across variable object count and different scene configurations, and infer latent properties of objects such as mass.

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。

↑