Search

Reservoir computing systems are promising for application in bio-inspired neuromorphic networks as they allow the considerable reduction of training energy and time costs as well as an overall system complexity. Conductive three-dimensional structures with the ability of reversible resistive switching are intensively developed to be applied in such systems. Nonwoven conductive materials, due to their stochasticity, flexibility and possibility of large-scale production, seem promising for this task. In this work, fabrication of a conductive 3D material by polyaniline synthesis on a polyamide-6 nonwoven matrix was shown. An organic stochastic device with a prospective to be used in reservoir computing systems with multiple inputs was created based on this material. The device demonstrates different responses (output current) when different combinations of voltage pulses are applied to the inputs. The approach is tested in handwritten digit image classification task in simulation with the overall accuracy exceeding 96%. This approach is beneficial for processing multiple data flows within a single reservoir device. [ABSTRACT FROM AUTHOR]

Nowadays, neuromorphic systems based on memristors are considered promising approaches to the hardware realization of artificial intelligence systems with efficient information processing. However, a major bottleneck in the physical implementation of these systems is the strong dependence of their performance on the unavoidable variations (cycle‐to‐cycle, c2c, or device‐to‐device, d2d) of memristive devices. Recently, reservoir computing (RC) and spiking neuromorphic systems (SNSs) are separately proposed as valuable options to partially mitigate this problem. Herein, both approaches are combined to create a fully organic system based on 1) volatile polyaniline memristive devices for the reservoir layer and 2) nonvolatile parylene memristors for the SNS readout layer. This combination provides a simpler SNS training procedure compared with the formal neural networks and results in greater robustness to device variability, while ensuring the extraction and encoding of the input critical features (performed by the polyaniline reservoir) and the analysis and classification performed by the SNS layer. Furthermore, the spatiotemporal pattern recognition of the system brings us closer to the implementation of efficient and reliable brain‐inspired computing systems built with partially unreliable analog elements. [ABSTRACT FROM AUTHOR]

One of the remarkable features of the emerging neuromorphic systems is the ability of implementing in‐memory computing which is demonstrated using memristors to realize both memory and computation functionalities within a single element. However, biological neural systems exhibit many other outstanding computing capabilities, among which one is the sensitivity to temporal parameters of neural activity. The identification and the realization of systems able to imitate this ability is still a very challenging perspective. Herein, polyaniline‐based organic memristive devices endowed with volatile resistive switching, complex temporal behaviors and capable of processing 4‐bit sequences of data with reliable separation of states are demonstrated. Thanks to this ability, such devices can be key elements in a reservoir layer of a network to map high‐dimensional input signals to a lower‐dimensional feature space. Herein, it is demonstrated through simulations that this type of device could be a valuable element for the realization of a reservoir computing system for the classification of handwritten digits from MNIST dataset. The model suggests that the electrical properties of the polyaniline‐based organic memristive devices ensure the realization of a system able to correctly classify handwritten digits and to be tolerant to considerable overlapping of neighboring reservoir states. [ABSTRACT FROM AUTHOR]

Existing methods of neurorehabilitation include invasive or non-invasive stimulators that are usually simple digital generators with manually set parameters like pulse width, period, burst duration, and frequency of stimulation series. An obvious lack of adaptation capability of stimulators, as well as poor biocompatibility and high power consumption of prosthetic devices, highlights the need for medical usage of neuromorphic systems including memristive devices. The latter are electrical devices providing a wide range of complex synaptic functionality within a single element. In this study, we propose the memristive schematic capable of self-learning according to bio-plausible spike-timing-dependant plasticity to organize the electrical activity of the walking pattern generated by the central pattern generator. [ABSTRACT FROM AUTHOR]

Convolutional neural networks (CNNs) have been widely used in image recognition and processing tasks. Memristor-based CNNs accumulate the advantages of emerging memristive devices, such as nanometer critical dimensions, low power consumption, and functional similarity to biological synapses. Most studies on memristor-based CNNs use either software models of memristors for simulation analysis or full hardware CNN realization. Here, we propose a hybrid CNN, consisting of a hardware fixed pre-trained and explainable feature extractor and a trainable software classifier. The hardware part was realized on passive crossbar arrays of memristors based on nanocomposite (Co-Fe-B)x(LiNbO3)100−x structures. The constructed 2-kernel CNN was able to classify the binarized Fashion-MNIST dataset with ~ 84% accuracy. The performance of the hybrid CNN is comparable to the other reported memristor-based systems, while the number of trainable parameters for the hybrid CNN is substantially lower. Moreover, the hybrid CNN is robust to the variations in the memristive characteristics: dispersion of 20% leads to only a 3% accuracy decrease. The obtained results pave the way for the efficient and reliable realization of neural networks based on partially unreliable analog elements. [ABSTRACT FROM AUTHOR]

Currently, there is growing interest in wearable and biocompatible smart computing and information processing systems that are safe for the human body. Memristive devices are promising for solving such problems due to a number of their attractive properties, such as low power consumption, scalability, and the multilevel nature of resistive switching (plasticity). The multilevel plasticity allows memristors to emulate synapses in hardware neuromorphic computing systems (NCSs). The aim of this work was to study Cu/poly- p -xylylene(PPX)/Au memristive elements fabricated in the crossbar geometry. In developing the technology for manufacturing such samples, we took into account their characteristics, in particular stable and multilevel resistive switching (at least 10 different states) and low operating voltage (<2 V), suitable for NCSs. Experiments on cycle to cycle (C2C) switching of a single memristor and device to device (D2D) switching of several memristors have shown high reproducibility of resistive switching (RS) voltages. Based on the obtained memristors, a formal hardware neuromorphic network was created that can be trained to classify simple patterns. [ABSTRACT FROM AUTHOR]

Organic memristive devices are promising elements to be used in memory tasks, neuromorphic systems, and bioelectronics as hardware analogs of synapses. Among other materials, small molecules are considered in this field as candidates with more adjustable properties and fine chemical structures. Herein, the first organic memristive device based on benzothieno[3,2‐b][1]‐benzothiophene (BTBT) siloxane dimer is reported. The device resistance programming by voltage pulses with various control parameters is demonstrated: amplitude, frequency, and pulse duty cycle. It is proved that the device has at least four different resistive states in each case. In addition, the device shows 2.5 × 104 cycles of sustainable operation in the endurance test with ROFF/RON ratio exceeding 5. [ABSTRACT FROM AUTHOR]

Spiking neuromorphic networks (SNNs) are bio-inspired artificial systems capable of unsupervised learning and promising candidates to mimic biological neural systems in efficient solution of cognitive tasks. Most SNNs are based on local learning rules, such as bio-like spike-time-dependent plasticity (STDP). In this paper, we report a significantly improved timescale of STDP for polyaniline-based memristive microdevices. We have used this result to show the possibility of associative learning with an unsupervised STDP-like mechanism of a simple SNN. The dependence of the required number of learning cycles on the pulse length was found: the longer the training pulse, the smaller the number of epochs the system needs to learn the associative rule. But the total training time remained nearly constant regardless of the pulse length. This study will be helpful in designing more sophisticated bio-plausible neuromorphic systems based on organic memristors. [ABSTRACT FROM AUTHOR]

Organic memristive devices are of great interest nowadays due to their low costs, flexibility, and prospects of use in artificial neuromorphic systems. Polyaniline (PANI) is a conductive and electrochemically active polymer and a promising material for redox‐gated memristive devices. In this work, the electrochromic properties of PANI are utilized for spatially resolved mapping of electrochemical reactions in the active zone of a model PANI‐based device. The observed nonuniform propagation of the reaction front defines the difference in the kinetics of the device while switching from the low‐ to high‐resistance state and vice versa. This study both reveals the mechanism of resistive switching in redox‐gated memristive elements and offers a novel nondestructive method for monitoring the resistive state of PANI‐based memristive devices. [ABSTRACT FROM AUTHOR]

In hardware neuromorphic systems (NSs), memristors are used as synaptic connections. In such systems, spike‐timing‐dependent plasticity (STDP) is a promising local learning rule. Herein, STDP is studied in a system composed of a pair of hardware or software neurons connected by a (CoFeB)x(LiNbO3)100−x nanocomposite‐based memristor. The dopamine‐like modulation of memristor‐based STDP is implemented simply by the change in the polarity of spikes generated by artificial neurons operating in the inhibitory or excitatory mode. This modulation method is shown to be compatible with hardware neurons, different spike shapes, and is used in spiking NSs with bioinspired dopamine‐like reinforcement learning. [ABSTRACT FROM AUTHOR]

Here we provide a perspective concept of neurohybrid memristive chip based on the combination of living neural networks cultivated in microfluidic/microelectrode system, metal-oxide memristive devices or arrays integrated with mixed-signal CMOS layer to control the analog memristive circuits, process the decoded information, and arrange a feedback stimulation of biological culture as parts of a bidirectional neurointerface. Our main focus is on the state-of-the-art approaches for cultivation and spatial ordering of the network of dissociated hippocampal neuron cells, fabrication of a large-scale cross-bar array of memristive devices tailored using device engineering, resistive state programming, or non-linear dynamics, as well as hardware implementation of spiking neural networks (SNNs) based on the arrays of memristive devices and integrated CMOS electronics. The concept represents an example of a brain-on-chip system belonging to a more general class of memristive neurohybrid systems for a new-generation robotics, artificial intelligence, and personalized medicine, discussed in the framework of the proposed roadmap for the next decade period. [ABSTRACT FROM AUTHOR]

Spiking neural networks (SNNs) are believed to be highly computationally and energy efficient for specific neurochip hardware real-time solutions. However, there is a lack of learning algorithms for complex SNNs with recurrent connections, comparable in efficiency with back-propagation techniques and capable of unsupervised training. Here we suppose that each neuron in a biological neural network tends to maximize its activity in competition with other neurons, and put this principle at the basis of a new SNN learning algorithm. In such a way, a spiking network with the learned feed-forward, reciprocal and intralayer inhibitory connections, is introduced to the MNIST database digit recognition. It has been demonstrated that this SNN can be trained without a teacher, after a short supervised initialization of weights by the same algorithm. Also, it has been shown that neurons are grouped into families of hierarchical structures, corresponding to different digit classes and their associations. This property is expected to be useful to reduce the number of layers in deep neural networks and modeling the formation of various functional structures in a biological nervous system. Comparison of the learning properties of the suggested algorithm, with those of the Sparse Distributed Representation approach shows similarity in coding but also some advantages of the former. The basic principle of the proposed algorithm is believed to be practically applicable to the construction of much more complicated and diverse task solving SNNs. We refer to this new approach as "Family-Engaged Execution and Learning of Induced Neuron Groups," or FEELING. [ABSTRACT FROM AUTHOR]

The problem with training spiking neural networks (SNNs) is relevant due to the ultra-low power consumption these networks could exhibit when implemented in neuromorphic hardware. The ongoing progress in the fabrication of memristors, a prospective basis for analogue synapses, gives relevance to studying the possibility of SNN learning on the base of synaptic plasticity models, obtained by fitting the experimental measurements of the memristor conductance change. The dynamics of memristor conductances is (necessarily) nonlinear, because conductance changes depend on the spike timings, which neurons emit in an all-or-none fashion. The ability to solve classification tasks was previously shown for spiking network models based on the bio-inspired local learning mechanism of spike-timing-dependent plasticity (STDP), as well as with the plasticity that models the conductance change of nanocomposite (NC) memristors. Input data were presented to the network encoded into the intensities of Poisson input spike sequences. This work considers another approach for encoding input data into input spike sequences presented to the network: temporal encoding, in which an input vector is transformed into relative timing of individual input spikes. Since temporal encoding uses fewer input spikes, the processing of each input vector by the network can be faster and more energy-efficient. The aim of the current work is to show the applicability of temporal encoding to training spiking networks with three synaptic plasticity models: STDP, NC memristor approximation, and PPX memristor approximation. We assess the accuracy of the proposed approach on several benchmark classification tasks: Fisher's Iris, Wisconsin breast cancer, and the pole balancing task (CartPole). The accuracies achieved by SNN with memristor plasticity and conventional STDP are comparable and are on par with classic machine learning approaches. [ABSTRACT FROM AUTHOR]

Parylene is a widely used polymer possessing such advantages as low cost and safety for the human body. Recently, several studies have been conducted showing that parylene can be used as a dielectric layer of memristors — new circuit design elements that are promising for the implementation of hardware neural networks. However, the mechanism of resistive switching of parylene-based memristors remains unclear. In this paper, we report the result of a comprehensive study of this mechanism for Metal/Parylene/ITO sandwich memristive devices. The obtained results clearly show that the origin of resistive switching in the devices is the formation of conductive metal bridges (filaments) from the top electrode (Cu, Ag or Al) to the bottom one (ITO). And furthermore, conductance quantization effect with both integer and half-integer multiples of the quantum of conductance G 0 = 2 e 2/ h has been observed in the samples, which also confirms the chosen switching model, and can be useful in the development of multilevel data memory cells. Image 1 • Resistive switching in Metal/Parylene/ITO memristors is driven by the ECM mechanism. • Devices exhibit quantized conductance, including half-integer multiples of G 0 = 2 e 2/ h. • Parylene-based memristors are promising building blocks of neuromorphic systems. [ABSTRACT FROM AUTHOR]

Memristive devices, known for their nonvolatile resistive switching, are promising components for next-generation neuromorphic computing systems, which mimic the brain's neural architecture. Specifically, these devices are well-suited for functioning as artificial synapses due to their analogue tunability and low energy consumption. However, the improvement of their performance and reliability remains a pressing challenge. In this study, we report the development and comprehensive characterization of memristive devices based on a parylene-MoO x (PPX-Mo) nanocomposite layer, which exhibit improved characteristics over their parylene-based counterparts: lower switching voltage and energy, smaller dispersion, and better resistive plasticity. A robust statistical analysis identified the optimal synthesis parameters for these devices, providing valuable insights for future device optimization. The most probable resistive switching mechanism of the devices is proposed. By successfully integrating these memristors into a neuromorphic computing model and showcasing their scalability in crossbar geometry, we demonstrate their potential as functional artificial synapses. The results obtained from this study can be useful for the development of hardware-brain-inspired computational systems.