Neuromorphic learning with Mott insulator NiO-猫眼课题宝

DOI：

10.1073/pnas.2017239118

发表时间：

2021-09

期刊：

Proceedings of the National Academy of Sciences

影响因子：

通讯作者：

中科院分区：

其他

文献类型：

作者：

Zhen Zhang;Sandip Mondal;S. Mandal;Jason M. Allred;N. A. Aghamiri;A. Fali;Zhan Zhang;Hua Zhou;Hui Cao;F. Rodolakis;J. McChesney;Qi Wang;Yifei Sun;Y. Abate;K. Roy;K. Rabe;S. Ramanathan

神经形态计算需要在合成物质中模拟动物学习。具有高度可调电子结构和对环境刺激的动态响应的材料特别适合于这项任务。在这里，我们展示了普遍的学习特性，如在一个原型的量子材料，NiO的习惯化和敏化。在氧气、臭氧和光等刺激下，原子缺陷的浓度可以被可逆地调节，从而导致模拟非联想学习的电导率变化。材料行为激发了神经网络中无监督学习的新算法，并为Mott绝缘体在人工智能中的应用开辟了新的方向。习惯化和敏感化（非联想学习）是生物体中最基本的学习和记忆行为形式之一，能够在动态环境中适应和学习。在固态中模拟自然界中发现的智能的这些特征可以作为人工神经网络算法模拟的灵感，并在神经形态计算中有潜在的用途。在这里，我们展示了非联想学习与一个典型的莫特绝缘体，氧化镍（NiO），在各种外部刺激在室温及以上。类似于生物物种，如Aaplasia，NiO的习惯化和敏化具有依赖于刺激之间的强度和时间间隔的时间依赖性可塑性。实验方法和第一性原理计算的结合揭示了NiO的这种学习行为是由于其缺陷和电子结构的动态调制。一个人工神经网络模型的启发，这种非联想学习模拟显示的优势，无监督聚类任务的准确性和减少灾难性的干扰，这可能有助于缓解稳定性，可塑性的困境。因此，莫特绝缘体可以作为构建块来检查生物学中注意到的学习行为，并激发人工智能的新学习算法。

Significance Neuromorphic computing requires emulation of animal learning in synthetic matter. Materials with highly tunable electronic structures and a dynamical response to environmental stimuli are particularly suited for this task. Here, we demonstrate universal learning characteristics such as habituation and sensitization in a prototypical quantum material, NiO. With stimuli such as oxygen, ozone, and light, the concentration of atomic defects can be modulated reversibly, resulting in changes to electrical conductivity that mimic nonassociative learning. The material behavior inspires new algorithms for unsupervised learning in neural networks and opens up new directions for use of Mott insulators in artificial intelligence. Habituation and sensitization (nonassociative learning) are among the most fundamental forms of learning and memory behavior present in organisms that enable adaptation and learning in dynamic environments. Emulating such features of intelligence found in nature in the solid state can serve as inspiration for algorithmic simulations in artificial neural networks and potential use in neuromorphic computing. Here, we demonstrate nonassociative learning with a prototypical Mott insulator, nickel oxide (NiO), under a variety of external stimuli at and above room temperature. Similar to biological species such as Aplysia, habituation and sensitization of NiO possess time-dependent plasticity relying on both strength and time interval between stimuli. A combination of experimental approaches and first-principles calculations reveals that such learning behavior of NiO results from dynamic modulation of its defect and electronic structure. An artificial neural network model inspired by such nonassociative learning is simulated to show advantages for an unsupervised clustering task in accuracy and reducing catastrophic interference, which could help mitigate the stability–plasticity dilemma. Mott insulators can therefore serve as building blocks to examine learning behavior noted in biology and inspire new learning algorithms for artificial intelligence.

Neuromorphic learning with Mott insulator NiO

文献摘要

被引文献

相似文献