Synaptic Weight Changes Research Articles

Lithography Processable Ta2O5 Barrier-Layered Chitosan Electric Double Layer Synaptic Transistors.

We proposed a synaptic transistor gated using a Ta2O5 barrier-layered organic chitosan electric double layer (EDL) applicable to a micro-neural architecture system. In most of the previous studies, a single layer of chitosan electrolyte was unable to perform lithography processes due to poor mechanical/chemical resistance. To overcome this limitation, we laminated a high-k Ta2O5 thin film on chitosan electrolyte to ensure high mechanical/chemical stability to perform a lithographic process for micropattern formation. Artificial synaptic behaviors were realized by protonic mobile ion polarization in chitosan electrolytes. In addition, neuroplasticity modulation in the amorphous In–Ga–Zn-oxide (a-IGZO) channel was implemented by presynaptic stimulation. We also demonstrated synaptic weight changes through proton polarization, excitatory postsynaptic current modulations, and paired-pulse facilitation. According to the presynaptic stimulations, the magnitude of mobile proton polarization and the amount of weight change were quantified. Subsequently, the stable conductance modulation through repetitive potential and depression pulse was confirmed. Finally, we consider that proposed synaptic transistor is suitable for advanced micro-neural architecture because it overcomes the instability caused when using a single organic chitosan layer.

Emulation of Synaptic Scaling Based on MoS2 Neuristor for Self‐Adaptative Neuromorphic Computing

AbstractRecent studies indicate that synaptic scaling is a vital mechanism to solve instability risks brought by the positive feedback of synaptic weight change related with standalone Hebbian plasticity. There are two kinds of synaptic scaling in the neural network, including local scaling and global scaling, both important for stabilizing the neural function. In this paper, for the first time, local synaptic scaling is emulated based on the MoS2 neuristor. The first‐principle calculation reveals that synaptic scaling achieved by the neuristor is associated with an internal residual Li+‐related weak dynamical process. Experimental results show the potential of achieving global synaptic scaling by the same device. Moreover, inspired by the synaptic scaling in the human brain, a new method of weight mapping called weight scaling mapping (WSM) is proposed to improve the stability of an artificial neural network (ANN). The simulation results indicate that WSM can improve the accuracy and anti‐noise ability of the network compared with the traditional mapping method. These findings provide new insight into bionic research and help advance the construction of stable neuromorphic systems.

Noise suppression ability and its mechanism analysis of scale-free spiking neural network under white Gaussian noise.

With the continuous improvement of automation and informatization, the electromagnetic environment has become increasingly complex. Traditional protection methods for electronic systems are facing with serious challenges. Biological nervous system has the self-adaptive advantages under the regulation of the nervous system. It is necessary to explore a new thought on electromagnetic protection by drawing from the self-adaptive advantage of the biological nervous system. In this study, the scale-free spiking neural network (SFSNN) is constructed, in which the Izhikevich neuron model is employed as a node, and the synaptic plasticity model including excitatory and inhibitory synapses is employed as an edge. Under white Gaussian noise, the noise suppression abilities of the SFSNNs with the high average clustering coefficient (ACC) and the SFSNNs with the low ACC are studied comparatively. The noise suppression mechanism of the SFSNN is explored. The experiment results demonstrate that the following. (1) The SFSNN has a certain degree of noise suppression ability, and the SFSNNs with the high ACC have higher noise suppression performance than the SFSNNs with the low ACC. (2) The neural information processing of the SFSNN is the linkage effect of dynamic changes in neuron firing, synaptic weight and topological characteristics. (3) The synaptic plasticity is the intrinsic factor of the noise suppression ability of the SFSNN.

An activity-dependent local transport regulation via degradation and synthesis of KIF17 underlying cognitive flexibility.

Synaptic weight changes among postsynaptic densities within a single dendrite are regulated by the balance between localized protein degradation and synthesis. However, the molecular mechanism via these opposing regulatory processes is still elusive. Here, we showed that the molecular motor KIF17 was locally degraded and synthesized in an N-methyl-d-aspartate receptor (NMDAR)-mediated activity-dependent manner. Accompanied by the degradation of KIF17, its transport was temporarily dampened in dendrites. We also observed that activity-dependent local KIF17 synthesis driven by its 3' untranslated region (3'UTR) occurred at dendritic shafts, and the newly synthesized KIF17 moved along the dendrites. Furthermore, hippocampus-specific deletion of Kif17 3'UTR disrupted KIF17 synthesis induced by fear memory retrieval, leading to impairment in extinction of fear memory. These results indicate that the regulation of the KIF17 transport is driven by the single dendrite-restricted cycle of degradation and synthesis that underlies cognitive flexibility.

Floating-gate photosensitive synaptic transistors with tunable functions for neuromorphic computing

Synaptic devices that merge memory and processing functions into one unit have broad application potentials in neuromorphic computing, soft robots, and human-machine interfaces. However, most previously reported synaptic devices exhibit fixed performance once been fabricated, which limits their application in diverse scenarios. Here, we report floating-gate photosensitive synaptic transistors with charge-trapping perovskite quantum dots (PQDs) and atomic layer deposited (ALD) Al2O3 tunneling layers, which exhibit typical synaptic behaviors including excitatory postsynaptic current (EPSC), pair-pulse facilitation and dynamic filtering characteristics under both electrical or optical signal stimulation. Further, the combination of the high-quality Al2O3 tuning layer and highly photosensitive PQDs charge-trapping layer provides the devices with extensively tunable synaptic performance under optical and electrical co-modulation. Applying light during electrical modulation can significantly improve both the synaptic weight changes and the non-linearity of weight updates, while the memory effect under light modulation can be obviously adjusted by the gate voltage. The pattern learning and forgetting processes for “0” and “1” with different synaptic weights and memory times are further demonstrated in the device array. Overall, this work provides synaptic devices with tunable functions for building complex and robust artificial neural networks.

Neuro-Transistor Based on UV-Treated Charge Trapping in MoTe2 for Artificial Synaptic Features.

The diversity of brain functions depend on the release of neurotransmitters in chemical synapses. The back gated three terminal field effect transistors (FETs) are auspicious candidates for the emulation of biological functions to recognize the proficient neuromorphic computing systems. In order to encourage the hysteresis loops, we treated the bottom side of MoTe2 flake with deep ultraviolet light in ambient conditions. Here, we modulate the short-term and long-term memory effects due to the trapping and de-trapping of electron events in few layers of a MoTe2 transistor. However, MoTe2 FETs are investigated to reveal the time constants of electron trapping/de-trapping while applying the gate-voltage pulses. Our devices exploit the hysteresis effect in the transfer curves of MoTe2 FETs to explore the excitatory/inhibitory post-synaptic currents (EPSC/IPSC), long-term potentiation (LTP), long-term depression (LTD), spike timing/amplitude-dependent plasticity (STDP/SADP), and paired pulse facilitation (PPF). Further, the time constants for potentiation and depression is found to be 0.6 and 0.9 s, respectively which seems plausible for biological synapses. In addition, the change of synaptic weight in MoTe2 conductance is found to be 41% at negative gate pulse and 38% for positive gate pulse, respectively. Our findings can provide an essential role in the advancement of smart neuromorphic electronics.

Hf0.5Zr0.5O2-based ferroelectric memristor with multilevel storage potential and artificial synaptic plasticity

Memristors are designed to mimic the brain’s integrated functions of storage and computing, thus breaking through the von Neumann framework. However, the formation and breaking of the conductive filament inside a conventional memristor is unstable, which makes it difficult to realistically mimic the function of a biological synapse. This problem has become a main factor that hinders memristor applications. The ferroelectric memristor overcomes the shortcomings of the traditional memristor because its resistance variation depends on the polarization direction of the ferroelectric thin film. In this work, an Au/Hf0.5Zr0.5O2/p+-Si ferroelectric memristor is proposed, which is capable of achieving resistive switching characteristics. In particular, the proposed device realizes the stable characteristics of multi-level storage, which possesses the potential to be applied to multi-level storage. Through polarization, the resistance of the proposed memristor can be gradually modulated by flipping the ferroelectric domains. Additionally, a plurality of resistance states can be obtained in bidirectional continuous reversibility, which is similar to the changes in synaptic weights. Furthermore, the proposed memristor is able to successfully mimic biological synaptic functions such as long-term depression, long-term potentiation, paired-pulse facilitation, and spike-timing-dependent plasticity. Consequently, it constitutes a promising candidate for a breakthrough in the von Neumann framework.

Short-Term Memory Dynamics of TiN/Ti/TiO2/SiOx/Si Resistive Random Access Memory.

In this study, we investigated the synaptic functions of TiN/Ti/TiO2/SiOx/Si resistive random access memory for a neuromorphic computing system that can act as a substitute for the von-Neumann computing architecture. To process the data efficiently, it is necessary to coordinate the information that needs to be processed with short-term memory. In neural networks, short-term memory can play the role of retaining the response on temporary spikes for information filtering. In this study, the proposed complementary metal-oxide-semiconductor (CMOS)-compatible synaptic device mimics the potentiation and depression with varying pulse conditions similar to biological synapses in the nervous system. Short-term memory dynamics are demonstrated through pulse modulation at a set pulse voltage of −3.5 V and pulse width of 10 ms and paired-pulsed facilitation. Moreover, spike-timing-dependent plasticity with the change in synaptic weight is performed by the time difference between the pre- and postsynaptic neurons. The SiOx layer as a tunnel barrier on a Si substrate provides highly nonlinear current-voltage (I–V) characteristics in a low-resistance state, which is suitable for high-density synapse arrays. The results herein presented confirm the viability of implementing a CMOS-compatible neuromorphic chip.

Synaptic Plasticity in Cortical Inhibitory Neurons: What Mechanisms May Help to Balance Synaptic Weight Changes?

Inhibitory neurons play a fundamental role in the normal operation of neuronal networks. Diverse types of inhibitory neurons serve vital functions in cortical networks, such as balancing excitation and taming excessive activity, organizing neuronal activity in spatial and temporal patterns, and shaping response selectivity. Serving these, and a multitude of other functions effectively requires fine-tuning of inhibition, mediated by synaptic plasticity. Plasticity of inhibitory systems can be mediated by changes at inhibitory synapses and/or by changes at excitatory synapses at inhibitory neurons. In this review, we consider that latter locus: plasticity at excitatory synapses to inhibitory neurons. Despite the fact that plasticity of excitatory synaptic transmission to interneurons has been studied in much less detail than in pyramids and other excitatory cells, an abundance of forms and mechanisms of plasticity have been observed in interneurons. Specific requirements and rules for induction, while exhibiting a broad diversity, could correlate with distinct sources of excitatory inputs and distinct types of inhibitory neurons. One common requirement for the induction of plasticity is the rise of intracellular calcium, which could be mediated by a variety of ligand-gated, voltage-dependent, and intrinsic mechanisms. The majority of the investigated forms of plasticity can be classified as Hebbian-type associative plasticity. Hebbian-type learning rules mediate adaptive changes of synaptic transmission. However, these rules also introduce intrinsic positive feedback on synaptic weight changes, making plastic synapses and learning networks prone to runaway dynamics. Because real inhibitory neurons do not express runaway dynamics, additional plasticity mechanisms that counteract imbalances introduced by Hebbian-type rules must exist. We argue that weight-dependent heterosynaptic plasticity has a number of characteristics that make it an ideal candidate mechanism to achieve homeostatic regulation of synaptic weight changes at excitatory synapses to inhibitory neurons.

Nonvolatile Memory and Artificial Synaptic Characteristics in Thin‐Film Transistors with Atomic Layer Deposited HfOx Gate Insulator and ZnO Channel Layer

AbstractNonvolatile memory and synaptic characteristics in thin‐film transistors (TFTs) with HfOx gate insulator and ZnO channel are investigated for the application to nonvolatile memory and artificial synapse in neuromorphic systems. Nonvolatile change of drain current induced by modulated gate stack properties is demonstrated to be applicable to nonvolatile memory operation. It also emulates synaptic weight change for learning and memory functions in artificial synapses. The TFTs with HfOx and ZnO layers deposited by sputtering or atomic layer deposition (ALD) at low temperatures exhibit tunable drain current upon applying gate pulses, featuring analog, reversible, nonvolatile changes with respect to pulse amplitude, width, interval, and repetition number. However, the TFTs with HfOx and ZnO by ALD at high temperatures show negligible change. The structural and chemical analyses reveal similarities in defective nature of sputter‐deposited and low‐temperature ALD HfOx and ZnO layers, leading to analogous drain current modulation. Also, the results of temperature‐ and voltage polarity‐dependent drain current changes and capacitance changes verify that the drain current modulation is driven by oxygen ion migration associated with defective states of HfOx and ZnO layers. It demonstrates feasibility of application of ALD‐HfOx/ZnO TFTs to nonvolatile memory and artificial synapses using modulated gate stack properties.

Fractional order controllers increase the robustness of closed-loop deep brain stimulation systems.

We studied the effects of using fractional order proportional, integral, and derivative (PID) controllers in a closed-loop mathematical model of deep brain stimulation. The objective of the controller was to dampen oscillations from a neural network model of Parkinson's disease. We varied intrinsic parameters, such as the gain of the controller, and extrinsic variables, such as the excitability of the network. We found that in most cases, fractional order components increased the robustness of the model multi-fold to changes in the gains of the controller. Similarly, the controller could be set to a fixed set of gains and remain stable to a much larger range, than for the classical PID case, of changes in synaptic weights that otherwise would cause oscillatory activity. The increase in robustness is a consequence of the properties of fractional order derivatives that provide an intrinsic memory trace of past activity, which works as a negative feedback system. Fractional order PID controllers could provide a platform to develop stand-alone closed-loop deep brain stimulation systems.

Brain-like spatiotemporal information processing with nanosized second-order synaptic emulators; “solid-state memory visualizer”

Abstract Emulating bio-counterpart artificial intelligence abilities at the hardware level requires high-density integration of artificial synapses. For that, not only nanosize artificial synapses with reliable multilevel functionality is needed but visualization of critical information processing within it is also essential. Here, we demonstrate nanosize second-order synaptic emulator and observed the dynamics of neuromorphic spatiotemporal information processing using local probe force microscopy. Particularly, the device shows stable analog hysteresis loop opening and multilevel memory storage, which was confirmed by current mapping. All versatile and necessary synaptic features like shot-, long-term memory, and corresponding dynamics, long term potentiation, and depression and pulse pair facilitation are demonstrated. Further, both Bienenstock, Cooper, and Munro learning rules e.g., frequency-dependent synaptic weight change and sliding threshold characteristics ‒ spike rate-dependent plasticity are confirmed. Based on Kelvin probe force microscopy measurements, the observed results are quantitatively explained as a dynamic of charge trapping/detrapping. Benefited from high-density integration and multilevel data storage along with processing, our approach exhibit an attractive future to build advance neuromorphic circuits at nanoscale.

Artificial synaptic transistors based on Schottky barrier height modulation using reduced graphene oxides

Development of artificial synapses is essential for highly-efficient brain-inspired neuromorphic computing. To achieve the high-performance artificial synapses, gradual change in synaptic weight with linear and symmetric forms is required. Here, we propose artificial synapses, in which synaptic weight is changed gradually by use of gate bias to modulate the Schottky barrier height between reduced graphene oxide and oxide semiconductor. This approach enables linear and symmetric change in synaptic weight. Also, an ion gel is used as the gate electrolyte, which can reduce the gate voltage required for modulation of Schottky barrier height. The fabricated artificial synapses show essential synaptic functions such as excitatory postsynaptic current, paired-pulse facilitation, and potentiation/depression. These results demonstrate that the devices that exploit modulation of Schottky barrier height can be applied to artificial synapses in advanced neuromorphic computing.

Li‐Ion Doping as a Strategy to Modulate the Electrical‐Double‐Layer for Improved Memory and Learning Behavior of Synapse Transistor Based on Fully Aqueous‐Solution‐Processed In2O3/AlLiO Film

AbstractNeuromorphic computing systems shows great potential for use in artificial intelligence. In this field, the simulation of synaptic behavior by electronic devices is considered as the first step for hardware implementation. However, solid‐state oxide synapse transistors still remain largely unexplored. An environmentally friendly aqueous route is adopted for oxide film fabrication. Li is doped in Al2O3 to promote the formation of electrical double layers (EDLs). An aqueous‐solution‐processed In2O3/AlLiO thin‐film transistor is fabricated and exhibits excellent electrical performance with a μ of 24.53 cm2 V−1 s−1 and a small VTH of 0.63 V. The device shows excellent short‐term memory (STM) and long‐term memory (LTM) characteristic response to electrical stimulus. By modulating the presynaptic pulse, the device exhibits a large change in synaptic weight (the maximum value of Δw is larger than 1000%). Conversion from STM to LTM is successfully stimulated. After repeated stimulation, the excitatory synaptic current (EPSC) can maintain a high‐level state for more than 100 s. Addtionally, high‐pass filtering and dynamic filtering characteristics are achieved and studied. Finally, Pavlov's dog experiments are conducted with this device to emulate the behavior of associative learning.

Optimality of sparse olfactory representations is not affected by network plasticity.

The neural representation of a stimulus is repeatedly transformed as it moves from the sensory periphery to deeper layers of the nervous system. Sparsening transformations are thought to increase the separation between similar representations, encode stimuli with great specificity, maximize storage capacity of associative memories, and provide an energy efficient instantiation of information in neural circuits. In the insect olfactory system, odors are initially represented in the periphery as a combinatorial code with relatively simple temporal dynamics. Subsequently, in the antennal lobe this representation is transformed into a dense and complex spatiotemporal activity pattern. Next, in the mushroom body Kenyon cells (KCs), the representation is dramatically sparsened. Finally, in mushroom body output neurons (MBONs), the representation takes on a new dense spatiotemporal format. Here, we develop a computational model to simulate this chain of olfactory processing from the receptor neurons to MBONs. We demonstrate that representations of similar odorants are maximally separated, measured by the distance between the corresponding MBON activity vectors, when KC responses are sparse. Sparseness is maintained across variations in odor concentration by adjusting the feedback inhibition that KCs receive from an inhibitory neuron, the Giant GABAergic neuron. Different odor concentrations require different strength and timing of feedback inhibition for optimal processing. Importantly, as observed in vivo, the KC–MBON synapse is highly plastic, and, therefore, changes in synaptic strength after learning can change the balance of excitation and inhibition, potentially leading to changes in the distance between MBON activity vectors of two odorants for the same level of KC population sparseness. Thus, what is an optimal degree of sparseness before odor learning, could be rendered sub–optimal post learning. Here, we show, however, that synaptic weight changes caused by spike timing dependent plasticity increase the distance between the odor representations from the perspective of MBONs. A level of sparseness that was optimal before learning remains optimal post-learning.

Multi-level anomalous Hall resistance in a single Hall cross for the applications of neuromorphic device

We demonstrate the process of obtaining memristive multi-states Hall resistance (RH) change in a single Hall cross (SHC) structure. Otherwise, the working mechanism successfully mimics the behavior of biological neural systems. The motion of domain wall (DW) in the SHC was used to control the ascend (or descend) of the RH amplitude. The primary synaptic functions such as long-term potentiation (LTP), long-term depression (LTD), and spike-time-dependent plasticity (STDP) could then be emulated by regulating RH. Applied programmable magnetic field pulses are in varying conditions such as intensity and duration to adjust RH. These results show that analog readings of DW movement can be closely resembled with the change of synaptic weight and have great potentials for bioinspired neuromorphic computing.

Anti-Injury Function of Complex Spiking Neural Networks Under Random Attack and Its Mechanism Analysis

The biological brain has self-adaptive ability through neural information processing and regulation. Drawing from the advantage of the biological brain, it is significant to research the robustness of artificial neural network (ANN) based on brain-like intelligence. In this study, based on the Izhikevich neuron model and the synaptic plasticity model which contains excitatory and inhibitory synapses, a spiking neural network (SNN) with small-world topology and a SNN with scale-free topology are constructed. The anti-injury function of two complex SNNs (CSNN) under random attack is comparatively analyzed. On this basis, the information processing of CSNN under attack is further discussed, and the anti-injury mechanism of CSNN is explored based on the synaptic plasticity. The experimental results show that: (1) scale-free SNN (SFSNN) has better performance than small-world SNN (SWSNN) in the anti-injury ability under random attacks. (2) The information processing of CSNN under random attacks is clarified by the linkage effects of dynamic changes in neuron firing, synaptic weight, and topological characteristics. (3) The anti-injury ability of CSNNs is closely related to the dynamic evolution of synaptic weight, which implies the dynamic regulation of synaptic plasticity is the intrinsic factor of the anti-injury function of CSNNs. This study lays a theoretical foundation for the application of brain-like intelligence with adaptive fault-tolerance.

Efficient implementation of synaptic learning rules for neuromorphic computing based on plasma-treated ZnO nanowire memristors

Nanomaterial-based memristors with analog resistive switching properties are used in the study of electronic synapses, providing information on both nanoscale device physics and low-power neuromorphic computing applications. Here, a memristor based on individual ZnO nanowires is prepared to study synaptic learning rules. Hebbian plasticity modulation is achieved with the co-application of pre- and post-synaptic spikes by tuning the temporal difference, spike frequency and voltage amplitude. Additionally, synaptic saturation is observed to stabilize the growth of synaptic weights. Plasma treatment of the memristors was performed to investigate its effects on synaptic plasticity and conductance modulation linearity during resistive switching. Plasma treatment allowed gradual conductance modulation of the memristor to be obtained, with improved conductance modulation linearity, suggesting that the memristor is capable of implementing synaptic plasticity to serve learning and memory. It was observed that the plasma treatment could also extend synaptic weight changes (Δw) to enhance learning capability and accelerate the learning speed of the electronic synapse, which might open up a route for modifying the characteristics of an electronic synapse. Synaptic learning and forgetting behavior are effectively simulated with re-learning of forgotten information at a much faster rate.

The Effect of Signaling Latencies and Node Refractory States on the Dynamics of Networks.

We describe the construction and theoretical analysis of a framework derived from canonical neurophysiological principles that model the competing dynamics of incident signals into nodes along directed edges in a network. The framework describes the dynamics between the offset in the latencies of propagating signals, which reflect the geometry of the edges and conduction velocities, and the internal refractory dynamics and processing times of the downstream node receiving the signals. This framework naturally extends to the construction of a perceptron model that takes into account such dynamic geometric considerations. We first describe the model in detail, culminating with the model of a geometric dynamic perceptron. We then derive upper and lower bounds for a notion of optimal efficient signaling between vertex pairs based on the structure of the framework. Efficient signaling in the context of the framework we develop here means that there needs to be a temporal match between the arrival time of the signals relative to how quickly nodes can internally process signals. These bounds reflect numerical constraints on the compensation of the timing of signaling events of upstream nodes attempting to activate downstream nodes they connect into that preserve this notion of efficiency. When a mismatch between signal arrival times and the internal states of activated nodes occurs, it can cause a breakdown in the signaling dynamics of the network. In contrast to essentially all of the current state of the art in machine learning, this work provides a theoretical foundation for machine learning and intelligence architectures based on the timing of node activations and their abilities to respond rather than necessary changes in synaptic weights. At the same time, the theoretical ideas we developed are guiding the discovery of experimentally testable new structure-function principles in the biological brain.

Corticostriatal synaptic weight evolution in a two-alternative forced choice task: a computational study

In natural environments, mammals can efficiently select actions based on noisy sensory signals and quickly adapt to unexpected outcomes to better exploit opportunities that arise in the future. Such feedback-based changes in behavior rely, in part, on long term plasticity within cortico-basal-ganglia-thalamic networks, driven by dopaminergic modulation of cortical inputs to the direct and indirect pathway neurons of the striatum. While the firing rates of striatal neurons have been shown to adapt across a range of feedback conditions, it remains difficult to directly assess the corticostriatal synaptic weight changes that contribute to these adaptive firing rates. In this work, we simulate the evolution of corticostriatal synaptic weights based on a spike timing-dependent plasticity rule driven by dopamine signaling that is induced by outcomes of actions in the context of a two-alternative forced choice task. Our results establish 1) that this plasticity model can successfully learn to select the most rewarding actions available, 2) that in the effective regime plasticity predominantly impacts direct pathway weights, evolving to drive action selection toward a more-rewarded action, and 3) that there can be coactivation of opposing populations within selected action channels, as observed experimentally. The model performance also agrees with the results of behavioral experiments carried out previously in human subjects using probabilistic reward paradigms.