清华大学类脑计算研究中心施路平教授领导的研究团队，在人工智能领域取得了重大突破。该团队在5月30日发表的成果中，成功研发出全球首款类脑互补视觉芯片“天眸芯”，这一成果登上了《自然》杂志的封面，标志着中国在类脑计算技术领域迈向了世界前沿。
在这一基础上，施路平团队近日再次推出了一种创新的神经形态计算架构——类脑神经计算模型Dendristor。Dendristor旨在模拟大脑中神经元之间的连接以及树突的树状结构，通过复制突触的组织和树突的树状结构，Dendristor旨在实现高能效的人工智能视觉感知能力。
Dendristor的创新之处在于其类脑形态树突网络计算模型，该模型由清华大学脑与智能实验室（THBI）的Eunhye Baek博士、宋森教授、Carlo Vittorio Cannistraci教授，以及清华大学精密仪器系的赵蓉教授和施路平教授共同完成。这一模型通过在类脑神经计算领域的新突破，展示了对大脑工作原理的深入理解，并有望在未来的AI技术发展中，提供更为高效、低功耗的视觉感知解决方案。
论文《Neuromorphic dendritic network computation with silent synapses for visual motion perception》于6月6日发表在《自然电子》杂志上，详细介绍了Dendristor模型的设计理念、实现方法以及在视觉运动感知领域的应用潜力。该研究不仅展示了清华大学在类脑计算领域的深厚积累，也为全球人工智能研究者提供了新的灵感和方向。
Dendristor的发布，标志着类脑计算技术在人工智能领域的应用又向前迈进了一大步。这一创新不仅提升了AI系统的能效，还有望在自动驾驶、机器人视觉、图像处理等多个领域带来革命性的变革。随着更多类脑计算模型的探索和应用，未来的人工智能系统将更加接近人类的思考方式，展现出更加智能、高效、节能的特点。
英语如下：
Headline: “Tsinghua University’s Innovation in Neural Computing Model, Achieving High-Efficiency Visual Perception”
Keywords: Brain-like chips, Dendristor architecture, High-efficiency AI
News Content: Headline: Tsinghua University’s Liu Ping Shi Team Unveils Brain-Like Neuromorphic Computing Model Dendristor, Achieving High-Efficiency AI Perception
A research team led by Professor Liu Ping Shi at the Tsinghua University’s Brain-like Computing Research Center has made a significant breakthrough in the field of artificial intelligence. The team, which presented its findings on May 30, successfully developed the world’s first brain-like complementary visual chip “Tianmoxin,” a feat that graced the cover of the prestigious journal ‘Nature,’ signaling China’s advancement in the realm of brain-like computing technologies.
Building on this achievement, Professor Shi’s team recently introduced a groundbreaking neuromorphic computing architecture – the brain-like neural computing model Dendristor. Designed to mimic the connections between neurons in the brain and the tree-like structure of dendrites, Dendristor aims to replicate the high-efficiency artificial intelligence visual perception capability through the duplication of synaptic organization and dendritic tree structures.
The innovation in Dendristor lies in its brain-like dendritic network computing model, developed by Dr. Eunhye Baek from the Tsinghua University Brain and Intelligence Laboratory (THBI), Prof. Song Sen, Prof. Carlo Vittorio Cannistraci, and Prof. Zhao Rong from Tsinghua University’s Precision Instrument Department, alongside Prof. Shi. This model showcases a deep understanding of the brain’s operational principles and is poised to provide more efficient, low-power visual perception solutions in future AI technologies.
The paper ‘Neuromorphic Dendritic Network Computation with Silent Synapses for Visual Motion Perception’ was published on June 6 in the journal ‘Nature Electronics,’ detailing the Dendristor model’s design concept, implementation methods, and potential applications in visual motion perception. This research not only highlights Tsinghua University’s extensive expertise in brain-like computing but also offers new inspiration and directions for global AI researchers.
The release of Dendristor marks a significant advancement in the application of brain-like computing technologies in the AI domain. This innovation not only enhances the energy efficiency of AI systems but also promises revolutionary changes in fields such as autonomous driving, robotic vision, and image processing. As more brain-like computing models are explored and applied, future AI systems will increasingly mimic human thinking, displaying more intelligent, efficient, and energy-saving characteristics.
【来源】https://www.jiqizhixin.com/articles/2024-07-12-2
Views: 3