Robust Spike-Based Continual Meta-Learning Improved by Restricted Minimum Error Entropy Criterion

Entropy, Journal Year: 2022, Volume and Issue: 24(4), P. 455 - 455

Published: March 25, 2022

The spiking neural network (SNN) is regarded as a promising candidate to deal with the great challenges presented by current machine learning techniques, including high energy consumption induced deep networks. However, there still gap between SNNs and online meta-learning performance of artificial Importantly, existing spike-based models do not target robust based on spatio-temporal dynamics superior theory. In this invited article, we propose novel framework minimum error entropy, called MeMEE, using entropy theory establish gradient-based scheme in recurrent SNN architecture. We examine various types tasks, autonomous navigation working memory test. experimental results show that proposed MeMEE model can effectively improve accuracy robustness performance. More importantly, emphasizes application modern information theoretic approach state-of-the-art algorithms. Therefore, paper, provide new perspectives for further integration advanced SNNs, which could be merit applied developments neuromorphic systems.

Language: Английский

---

Progress, Challenges, and Opportunities in Oxide Semiconductor Devices: A Key Building Block for Applications Ranging from Display Backplanes to 3D Integrated Semiconductor Chips

Advanced Materials, Journal Year: 2022, Volume and Issue: 35(43)

Published: July 21, 2022

As Si has faced physical limits on further scaling down, novel semiconducting materials such as 2D transition metal dichalcogenides and oxide semiconductors (OSs) have gained tremendous attention to continue the ever-demanding downscaling represented by Moore's law. Among them, OS is considered be most promising alternative material because it intriguing features modest mobility, extremely low off-current, great uniformity, low-temperature processibility with conventional complementary-metal-oxide-semiconductor-compatible methods. In practice, successfully replaced hydrogenated amorphous in high-end liquid crystal display devices now become a standard backplane electronic for organic light-emitting diode displays despite short time since their invention 2004. For implemented next-generation electronics back-end-of-line transistor applications monolithic 3D integration beyond applications, however, there still much room study, high immune short-channel effects, electrical contact properties, etc. This study reviews brief history of recent progress device from science physics point view. Simultaneously, remaining challenges opportunities use are discussed.

Language: Английский

---

Fully Printed Optoelectronic Synaptic Transistors Based on Quantum Dot–Metal Oxide Semiconductor Heterojunctions

ACS Nano, Journal Year: 2022, Volume and Issue: 16(6), P. 8651 - 8661

Published: April 22, 2022

Optoelectronic synaptic transistors with hybrid heterostructure channels have been extensively developed to construct artificial visual systems, inspired by the human system. However, optoelectronic taking full advantages of superior behaviors, low-cost processes, low-power consumption, and environmental benignity remained a challenge. Herein, we report fully printed, high-performance transistor based on heterostructures heavy-metal-free InP/ZnSe core/shell quantum dots (QDs) n-type SnO2 amorphous oxide semiconductors (AOSs). The elaborately designed heterojunction improves separation efficiency photoexcited charges, leading high photoresponsivity tunable weight changes. Under coordinated modulation electrical optical modes, important biological including excitatory postsynaptic current, short/long-term plasticity, paired-pulse facilitation, were demonstrated low power consumption (∼5.6 pJ per event). QD/SnO2 vision system illustrated significantly improved accuracy 91% in image recognition, compared that bare counterparts (58%). Combining outstanding characteristics both AOS materials structures, this work provides printable, low-cost, high-efficiency strategy achieve advanced synapses for neuromorphic electronics intelligence.

Language: Английский

---

Retina‐Inspired Artificial Synapses with Ultraviolet to Near‐Infrared Broadband Responses for Energy‐Efficient Neuromorphic Visual Systems

Advanced Functional Materials, Journal Year: 2023, Volume and Issue: 33(32)

Published: April 21, 2023

Abstract Neuromorphic visual system with image perception, memory, and preprocessing functions is expected to simulate basic features of the human retina. Organic optoelectronic synaptic transistors emulating biological synapses may be promising candidates for constructing neural morphological system. However, sensing wavelength range organic usually limits their potential in artificial multispectral perception. Here, retina‐inspired that present broadband responses covering ultraviolet, visible, near‐infrared regions are demonstrated, which leverage wide‐range photoresponsive charge trapping layer heterostructure formed between PbS quantum dots semiconductor. Simplified neuromorphic arrays developed comprehensive functions. Benefitting from flexibility semiconductor layers, a flexible array can fabricated, having an ultralow power consumption 0.55 fJ per event under low operating voltage −0.01 V. More significantly, accelerating effect observed wide even beyond perception system, due gate‐adjustable plasticity. These devices highly implementing systems increasing processing efficiency, promoting development vision.

Language: Английский

---

Self-powered high-sensitivity all-in-one vertical tribo-transistor device for multi-sensing-memory-computing

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: Dec. 23, 2022

Devices with sensing-memory-computing capability for the detection, recognition and memorization of real time sensory information could simplify data conversion, transmission, storage, operations between different blocks in conventional chips, which are invaluable sought-after to offer critical benefits accomplishing diverse functions, simple design, efficient computing simultaneously internet things (IOT) era. Here, we develop a self-powered vertical tribo-transistor (VTT) based on MXenes multi-sensing-memory-computing function multi-task emotion recognition, integrates triboelectric nanogenerator (TENG) transistor single device configuration organic field effect (VOFET). The tribo-potential is found be able tune ionic migration insulating layer Schottky barrier height at MXene/semiconductor interface, thus modulate conductive channel MXene drain electrode. Meanwhile, sensing sensitivity can significantly improved by 711 times over TENG device, VTT exhibits excellent function. Importantly, this function, multi-sensing integration multi-model constructed, improves accuracy up 94.05% reliability. This structure high sensitivity, efficiency accuracy, provides application prospects future human-mechanical interaction, IOT high-level intelligence.

Language: Английский

---

Progress of Materials and Devices for Neuromorphic Vision Sensors

Nano-Micro Letters, Journal Year: 2022, Volume and Issue: 14(1)

Published: Oct. 15, 2022

The latest developments in bio-inspired neuromorphic vision sensors can be summarized 3 keywords: smaller, faster, and smarter. (1) Smaller: Devices are becoming more compact by integrating previously separated components such as sensors, memory, processing units. As a prime example, the transition from traditional sensory computing to in-sensor has shown clear benefits, simpler circuitry, lower power consumption, less data redundancy. (2) Swifter: Owing nature of physics, smaller integrated devices detect, process, react input quickly. In addition, methods for sensing optical information using various materials (such oxide semiconductors) evolving. (3) Smarter: these two main research directions, we expect advanced applications adaptive collision nociceptive sensors. This review mainly focuses on recent progress, working mechanisms, image pre-processing techniques, features types based near-sensor methodologies.

Language: Английский

---

Tailoring neuroplasticity in flexible perovskite QDs-based optoelectronic synaptic transistors by dual modes modulation

Nano Energy, Journal Year: 2022, Volume and Issue: 95, P. 106987 - 106987

Published: Jan. 25, 2022

Language: Английский

---

Advanced Optoelectronic Devices for Neuromorphic Analog Based on Low‐Dimensional Semiconductors

Advanced Functional Materials, Journal Year: 2023, Volume and Issue: 33(15)

Published: Jan. 22, 2023

Abstract Neuromorphic systems can parallelize the perception and computation of information, making it possible to break through von Neumann bottleneck. engineering has been developed over a long period time based on Hebbian learning rules. The optoelectronic neuromorphic analog device combines advantages electricity optics, simulate biological visual system, which very strong development potential. Low‐dimensional materials play important role in field devices due their flexible bandgap tuning mechanism light‐matter coupling efficiency. This review introduces basic synaptic plasticity devices. According different number terminals, two‐terminal memristors, three‐terminal transistors artificial system are introduced from aspects action structure. Finally, prospect low‐dimensional is prospected.

Language: Английский

---

Emerging Iontronic Neural Devices for Neuromorphic Sensory Computing

Advanced Materials, Journal Year: 2023, Volume and Issue: 35(39)

Published: March 9, 2023

Living organisms have a very mysterious and powerful sensory computing system based on ion activity. Interestingly, studies iontronic devices in the past few years proposed promising platform for simulating sensing functions of living organisms, because: 1) can generate, store, transmit variety signals by adjusting concentration spatiotemporal distribution ions, which analogs to how brain performs intelligent alternating flux polarization; 2) through ionic-electronic coupling, bridge biosystem with electronics offer profound implications soft electronics; 3) diversity be designed recognize specific ions or molecules customizing charge selectivity, ionic conductivity capacitance adjusted respond external stimuli schemes, more difficult electron-based devices. This review provides comprehensive overview emerging neuromorphic devices, highlighting representative concepts both low-level high-level introducing important material device breakthroughs. Moreover, as means are discussed regarding pending challenges future directions.

Language: Английский

---

Oxide Semiconductor Memristor‐Based Optoelectronic Synaptic Devices With Quaternary Memory Storage

Advanced Electronic Materials, Journal Year: 2024, Volume and Issue: 10(7)

Published: March 11, 2024

Abstract A pioneering integration of oxide semiconductor memristors with optoelectronic features is presented, surpassing binary limitations to realize multi‐valued synaptic operations. Through Pt/Ga 2 O 3 /Pt memristors, their structural and electronic attributes via atomic force microscopy, X‐ray diffraction, photoelectron spectroscopy are explored. Demonstrating unipolar resistance switching remarkable endurance retention, the devices exhibit intricate light‐resistance correlations, yielding substantial photoelectric effects in distinct states. Investigating behaviors, potentiation, depression akin biological synapses unveiled, facilitating learning memory processes. The standout achievement lies attaining quaternary storage within a single device. Empirical data simulations validate this concept, showcasing potential for encoding sustaining multiple This innovation heralds transformative possibilities, emphasizing as gateway enhanced functions. In essence, work pioneers expanded capabilities.

Language: Английский

---