《科学》（20230915出版）一周论文导读

 
 
《科学》（20230915出版）一周论文导读  
 

 

编译 | 未玖

Science,  15 SEP 2023, VOL 381, ISSUE 6663

《科学》2023年9月15日，第381卷，6663期

  ?

人工智能Artificial Intelligence

Edge learning using a fully integrated neuro-inspired memristor chip

使用全集成类脑忆阻器芯片进行边缘学习

▲ 作者：Wenbin Zhang, Peng Yao, Bin Gao, Qi Liu, Dong Wu, Qingtian Zhang, et al.

▲ 链接：https://www.science.org/doi/full/10.1126/science.ade3483

▲ 摘要：

学习对于边缘智能设备适应不同的应用场景和用户非常重要。当前训练神经网络的技术需要在计算和存储单元之间移动大量数据，这阻碍了在边缘设备上实现学习。

研究组开发了一种全集成忆阻器芯片，提高了学习能力，降低了能耗。STELLAR架构中包括学习算法、硬件实现和并行电导调制策略的方案，是通过使用忆阻器交叉阵列来促进片上学习的通用方法，而与忆阻器器件类型无关。

在这项研究中执行的任务包括运动控制、图像分类和语音识别。

▲ Abstract：

Learning is highly important for edge intelligence devices to adapt to different application scenes and owners. Current technologies for training neural networks require moving massive amounts of data between computing and memory units, which hinders the implementation of learning on edge devices. We developed a fully integrated memristor chip with the improvement learning ability and low energy cost. The schemes in the STELLAR architecture, including its learning algorithm, hardware realization, and parallel conductance tuning scheme, are general approaches that facilitate on-chip learning by using a memristor crossbar array, regardless of the type of memristor device. Tasks executed in this study included motion control, image classification, and speech recognition.

Powerful, soft combustion actuators for insect-scale robots

用于昆虫级机器人的强大、软燃烧致动器

▲ 作者：Cameron A. Aubin, Ronald H. Heisser, Ofek Peretz, Julia Timko, Jacqueline Lo, E. Farrell Helbling, et al.

▲ 链接：https://www.science.org/doi/full/10.1126/science.adg5067

▲ 摘要：

昆虫相对于其较小的身体，拥有惊人的强大力量和耐力。昆虫级机器人尽管遵循相同的缩放定律，但与昆虫相比性能却大大下降，这是因为当前微致动器技术由低能量密度电源驱动，产生的力和/或位移都很小。使用高能量密度化学燃料为小型软致动器提供动力是一种潜在解决方案。

研究组展示了一个325毫克的软燃烧微致动器，可实现140%的位移，工作频率>100赫兹，并产生>9.5牛顿的力。利用这些致动器为一个昆虫大小的四足机器人提供动力，该机器人展示了各种步态模式、方向控制和22倍于其体重的有效载荷能力。这些特征使昆虫机器人能够在崎岖地形和障碍物上移动。

▲ Abstract：

Insects perform feats of strength and endurance that belie their small stature. Insect-scale robots—although subject to the same scaling laws—demonstrate reduced performance because existing microactuator technologies are driven by low–energy density power sources and produce small forces and/or displacements. The use of high–energy density chemical fuels to power small, soft actuators represents a possible solution. We demonstrate a 325-milligram soft combustion microactuator that can achieve displacements of 140%, operate at frequencies >100 hertz, and generate forces >9.5 newtons. With these actuators, we powered an insect-scale quadrupedal robot, which demonstrated a variety of gait patterns, directional control, and a payload capacity 22 times its body weight. These features enabled locomotion through uneven terrain and over obstacles.

医学Medicine

MEG3 activates necroptosis in human neuron xenografts modeling Alzheimer’s disease

在模拟阿尔茨海默病的人类神经元异种移植物中MEG3激活坏死性凋亡

▲ 作者：Sriram Balusu, Katrien Horré, Nicola Thrupp, Katleen Craessaerts, An Snellinx, Lutgarde Serneels, et al.

▲ 链接：https://www.science.org/doi/full/10.1126/science.abp9556

▲ 摘要：

神经元丢失是阿尔茨海默病（AD）的一个决定性特征，但其潜在机制尚不清楚。