A self-adaptive hardware with resistive switching synapses for experience-based neurocomputing

Abstract

Neurobiological systems continually interact with the surrounding environment to refine their behaviour toward the best possible reward. Achieving such learning by experience is one of the main challenges of artificial intelligence, but currently it is hindered by the lack of hardware capable of plastic adaptation. Here, we propose a bio-inspired recurrent neural network, mastered by a digital system on chip with resistive-switching synaptic arrays of memory devices, which exploits homeostatic Hebbian learning for improved efficiency. All the results are discussed experimentally and theoretically, proposing a conceptual framework for benchmarking the main outcomes in terms of accuracy and resilience. To test the proposed architecture for reinforcement learning tasks, we study the autonomous exploration of continually evolving environments and verify the results for the Mars rover navigation. We also show that, compared to conventional deep learning techniques, our in-memory hardware has the potential to achieve a significant boost in speed and power-saving.

Synergistic Effects of UV Irradiation on a Flexible Ag/NCQDs-MoS2/Ag/Kapton Resistive-Switching Memory Device

“…In today’s high-tech world, the demands on memory technology are becoming increasingly stringent, due to the rapid advancements in various data-related applications and information processing. − Numerous memory concepts have been studied to address the data storage challenges and explore novel memory technologies. A resistive-switching (RS) phenomenon in a capacitor-like pattern with memristor material stacking between two metal electrodes has garnered significant attention due to its promising applications in various domains, including neuromorphic computing and next-generation memory devices. − In two such terminal devices, the RS behavior is typically driven by various processes such as phase changes (thermally activated amorphous–crystalline transition), ferroelectric switching (lattice-polarization-dependent tunneling), tunnel magnetoresistance (spin-dependent tunnel resistance), and electrochemical reactions (namely, redox and ion migration). − Abrupt resistive switching (ARS) and gradual resistive switching (GRS) are intriguing phenomena observed in specific types of resistive random-access memory (ReRAM) devices. GRS, for instance, allows for precise control over the programming process.…”

N-Doped Graphene/MXene Nanocomposite as a Temperature-Adaptive Neuromorphic Memristor

“…An adaptive low-power mode post-warm-up operation is suggested to maintain synaptic learning behavior while minimizing energy consumption. 11,12 Research progresses on temperature-adaptive, low-power electronic synaptic memristors, enhancing performance and reliability for AI and neuromorphic computing. 13,14 Evidently, recent reports in the study of temperature adaptive amnesia have reported explorations from the material design, device physics, and application levels, suggesting that the high-temperature performance of amnesia is important and reliable in neuromorphic computation and that the development of it has some potential applications in the power electronics and aerospace industries.…”

“…In integrated circuit applications, erase and write processes lead to elevated device temperature and power consumption, impacting the reliability and lifespan. An adaptive low-power mode post-warm-up operation is suggested to maintain synaptic learning behavior while minimizing energy consumption. , Research progresses on temperature-adaptive, low-power electronic synaptic memristors, enhancing performance and reliability for AI and neuromorphic computing. , …”

Fabrication of ZnO-encapsulated porous alumina structure utilizing a two-step anodizing technique for memristor applications

“…[1][2][3][4][5][6][7] In realizing an optimum hardware system for NC, the utilization of a memristor as the basic building block, which resembles the synapse cell of the human brain, is being pursued extensively due to its simplicity and versatility. [8][9][10][11] Up to now, various compound materials such as AlO x , 12) As 2 S 3 , 13) CuAg, 14) CuO x , 15) Ge 2 Se 3 , 16) HfO x , 17) MoO x , 18) NiO x , 19) SiO x , 20,21) SnS 2 , 22) TaO x , 23) TiO x , 24) VO 2 , 25) Y 2 O 3 , [26][27][28][29] ZnO, 30) and others are being exploited as the core part of the resistive changing medium of the memristors. However, reproducibility and stability of the device characteristic based on the memristor is still a challenging issue since the control of defects formation, which influences the electrical properties of the materials used for the resistive changing medium, is quite critical.…”