小柯机器人

美国纽约大学Xiao-Jing Wang团队近期取得重要工作进展，他们研究发现神经群体子空间中的模式形成是灵活的感觉运动问题解决中learning-to-learn的基础。相关研究成果2023年4月6日在线发表于《自然—神经科学》杂志上。

据介绍，Learning-to-learn是在解决一系列类似问题的同时逐步加速学习，代表了神经科学和人工智能领域引起关注的知识获取的核心过程。

为了研究其潜在的大脑机制，研究人员在已知依赖于前额叶皮层的任意感觉运动映射上训练了一个递归神经网络模型。该网络显示出加速学习的指数时间进程。模式的神经基底出现在种群活动的低维子空间中，它在新问题中的重用通过限制连接权重的变化来促进学习。

总之，这一工作强调了向量场的权重驱动修改，它决定了递归网络和行为的种群轨迹。这种可塑性对于保存和重用所学习的模式尤其重要，尽管由于向学习新问题的过渡而导致向量场发生了不希望的变化，问题之间累积的变化说明了Learning-to-learn的动力学。

附：英文原文

Title: Schema formation in a neural population subspace underlies learning-to-learn in flexible sensorimotor problem-solving

Author: Goudar, Vishwa, Peysakhovich, Barbara, Freedman, David J., Buffalo, Elizabeth A., Wang, Xiao-Jing

Issue&Volume: 2023-04-06

Abstract: Learning-to-learn, a progressive speedup of learning while solving a series of similar problems, represents a core process of knowledge acquisition that draws attention in both neuroscience and artificial intelligence. To investigate its underlying brain mechanism, we trained a recurrent neural network model on arbitrary sensorimotor mappings known to depend on the prefrontal cortex. The network displayed an exponential time course of accelerated learning. The neural substrate of a schema emerges within a low-dimensional subspace of population activity; its reuse in new problems facilitates learning by limiting connection weight changes. Our work highlights the weight-driven modifications of the vector field, which determines the population trajectory of a recurrent network and behavior. Such plasticity is especially important for preserving and reusing the learned schema in spite of undesirable changes of the vector field due to the transition to learning a new problem; the accumulated changes across problems account for the learning-to-learn dynamics.

DOI: 10.1038/s41593-023-01293-9

Source: https://www.nature.com/articles/s41593-023-01293-9

Nature Neuroscience：《自然—神经科学》，创刊于1998年。隶属于施普林格·自然出版集团，最新IF：28.771
官方网址：https://www.nature.com/neuro/
投稿链接：https://mts-nn.nature.com/cgi-bin/main.plex