NIH Director's Pioneer Award

我建议构建NeuroGrid，这是一个专门的硬件平台，它将执行皮质级别的仿真，同时提供 类似软件的灵活性。最近在脑成像方面的突破提供了一个前所未有的机会来理解 大脑是如何工作的，对社会有着深远的影响。要了解大脑，我们必须 通过对大脑建模来解释这些丰富的观察结果，这是测试我们理解能力的唯一方法吗？ 因为用生物部分建造一个真正的大脑目前是不可行的。NeuroGrid将进行仿真(在 实时)通过60亿个突触连接的100万个神经元？这使得对垂直、 生物物理细节中的水平和自上而下的皮质相互作用。 我有能力使这一努力取得成果，我对生物医学的承诺从我过去的成就中可见一斑。 在过去的八年里，我的实验室设计了七个神经形态芯片， 神经系统--视网膜、耳蜗、耳蜗核、丘脑、海马体、视皮层和视网膜顶盖。 发展。为了追求如此多样化的项目，我与6名同事在 宾夕法尼亚大学S神经科学系；我们的工作是《科学美国人》的封面故事(2005年5月)。视皮层 芯片展示了模拟VLSI的潜力：模拟9216个神经元，速度是 最先进的。然而，这个芯片和其他六个芯片都不是可编程的。 Neurogrid将通过使用数字VLSI增强模拟VLSI来提供可编程性，这是一种混合模式方法 这结合了两个世界的最好之处。虽然在模型中包含生物物理细节可以提供联系 通过实验，可编程性支持复制操作、执行控制、基准测试 模型和探索机制。在不牺牲规模的情况下实现这两个关键功能将使 第一次有可能复制实验室动物在生物真实模型中执行的任务， 我将与两位神经生理学家(马修·达尔瓦博士和威廉博士)密切合作 纽瑟姆博士)。

I propose to build Neurogrid, a specialized hardware platform that will perform cortex-scale emulations while offering software-like flexibility. Recent breakthroughs in brain mapping present an unprecedented opportunity to understand how the brain works, with profound implications for society. To understand the brain, we have to interpret these richly growing observations by modeling the brain, the only way to test our understanding? since building a real brain out of biological parts is currently infeasible. Neurogrid will emulate (simulate in real-time) one million neurons connected by six billion synapses?making it possible to model vertical, horizontal and top-down cortical interactions in biophysical detail. My ability to bring this endeavor to fruition and my commitment to biomedicine is evident in my past accomplishments. Over the past eight years, my lab has designed seven neuromorphic chips that model seven neural systems?retina, cochlea, cochlear nucleus, thalamus, hippocampus, visual cortex, and retinotectal development. To pursue such diverse projects, I established productive collaborations with six colleagues in Penn?s Neuroscience Department; our work was a Scientific American cover story (May 2005). The visual cortex chip illustrates the potential of Analog VLSI: Emulating 9,216 neurons, it is 2,765 times faster than the state-of-the-art. However, neither this chip nor the other six is programmable. Neurogrid will provide programmability by augmenting Analog VLSI with Digital VLSI, a mixed-mode approach that combines the best of both worlds. While including biophysical detail in a model provides contact with experiment, programmability supports replicating manipulations, performing controls, benchmarking models, and exploring mechanism. Realizing these two critical functions without sacrificing scale will make it possible to replicate tasks laboratory animals perform in biologically realistic models for the first time, which I will do in close collaboration with two neurophysiologists (Matthew Dalva Ph.D. and William Newsome Ph.D.).