GCR: Synthetic Neurocomputers for Cognitive Information Processing

该项目汇集了材料科学家和电气工程师，他们用嵌入式电子和光电设备构建合成3D支架，神经科学家在3D支架上培养神经细胞，形成具有精确定义的3D拓扑结构和集成多模态信息接口的生物神经网络，化学家和化学工程师合成功能分子，用于控制神经细胞的放置，发育和活动，以及计算机科学家，他们操作神经计算机原型，并利用机器学习方法提取其信息编码和处理算法。该项目的目标是为大脑逆向工程的巨大挑战做出贡献，并基于培养的生物神经网络开辟新的计算范式，以推动机器学习和人工智能更上一层楼。该项目中追求的神经计算机原型采用生物神经元电路，这些神经元电路被设计成定义良好的3D拓扑结构，让人想起深层神经网络模型作为信息处理单元。电子和光电设备将与每个神经细胞集成，以基于电生理学、光遗传学和神经化学来管理和监测神经元和突触活动。然后，制造的神经计算机原型将用于执行各种学习和计算任务，如图像识别和空间导航。神经代码和学习算法将使用基于尖峰生成模型和递归神经网络模型的实验和模拟相结合来提取。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The project brings together material scientists and electrical engineers who build synthetic 3D scaffolds with embedded electronic and optoelectronic devices, neuroscientists who culture neural cells on the 3D scaffold to form biological neural networks with precisely defined 3D topology and integrated multimodal information interfaces, chemists and chemical engineers who synthesize functional molecules for controlling the neural cell placement, development, and activity, and computer scientists who operate the neurocomputer prototype and extract its information-coding and processing algorithms with machine-learning methods. The goals of the project are to contribute to the grand challenge of reverse engineering the brain and open up new computing paradigms based on cultured biological neural networks to propel machine learning and artificial intelligence to the next level. The neurocomputer prototype pursued in the project employs biological neuronal circuits engineered into well-defined 3D topologies reminiscent of deep-neural-network models as the information-processing units. Electronic and optoelectronic devices will be integrated with each neural cell to administer and monitor the neuronal and synaptic activities based on electrophysiology, optogenetics, and neurochemistry. The fabricated neurocomputer prototype will then be utilized to perform various learning and computing tasks such as image recognition and space navigation. Neural code and learning algorithms will be extracted using a combination of experiment and simulations based on spike generation models and recurrent neural network models. The results will help reveal how complex living neural networks function and provide a technologically transformative approach to information-processing machines.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.