Memristors are widely considered as promising elements for the efficient implementation of synaptic weights in artificial neural networks (ANNs) since they are resistors that keep memory of their previous conductive state. Whereas demonstrations of simple neural networks (e.g., a single-layer perceptron) based on memristors already exist, the implementation of more complicated networks is more challenging and has yet to be reported. In this study, we demonstrate linearly nonseparable combinational logic classification (XOR logic task) using a network implemented with CMOS-based neurons and organic memrisitive devices that constitutes the first step toward the realization of a double layer perceptron. We also show numerically the ability of such network to solve a principally analogue task which cannot be realized by digital devices. The obtained results prove the possibility to create a multilayer ANN based on memristive devices that paves the way for designing a more complex network such as the double layer perceptron.

Neuromorphic systems consisting of artificial neurons and memristive synapses could provide a much better performance and a significantly more energy-efficient approach to the implementation of different types of neural network algorithms than traditional hardware with the Von-Neumann architecture. However, the memristive weight adjustment in the formal neuromorphic networks by the standard back-propagation techniques suffers from poor device-to-device reproducibility. One of the most promising approaches to overcome this problem is to use local learning rules for spiking neuromorphic architectures which potentially could be adaptive to the variability issue mentioned above. Different kinds of local rules for learning spiking systems are mostly realized on a bio-inspired spike-time-dependent plasticity (STDP) mechanism, which is an improved type of classical Hebbian learning. Whereas the STDP-like mechanism has already been shown to emerge naturally in memristive devices, the demonstration of its self-adaptive learning property, potentially overcoming the variability problem, is more challenging and has yet to be reported. Here we experimentally demonstrate an STDP-based learning protocol that ensures self-adaptation of the memristor resistive states, after only a very few spikes, and makes the plasticity sensitive only to the input signal configuration, but neither to the initial state of the devices nor their device-to-device variability. Then, it is shown that the self-adaptive learning of a spiking neuron with memristive weights on rate-coded patterns could also be realized with hardware-based STDP rules. The experiments have been carried out with nanocomposite-based (Co ₄₀ Fe ₄₀ B ₂₀ ) _х (LiNbO _{3− y} ) _{100− х} memristive structures, but their results are believed to be applicable to a wide range of memristive devices. All the experimental data were supported and extended by numerical simulations. There is a hope that the obtained results pave the way for building up reliable spiking neuromorphic systems composed of partially unreliable analog memristive elements, with a more complex architecture and the capability of unsupervised learning.

Polyaniline (PANI) based memristive devices have emerged as promising candidates for hardware implementation of artificial synapses (the key components of neuromorphic systems) due to their high flexibility, low cost, solution processability, three-dimensional stacking capability, and biocompatibility. Here, we report on a way of the significant improvement of the switching rate and endurance of PANI-based memristive devices. The reduction of the PANI active channel dimension leads to the increase in the resistive switching rate by hundreds of times in comparison with the conventional one. The miniaturized memristive device was shown to be stable within at least 10⁴ cyclic switching events between high- and low-conductive states with a retention time of at least 10³ s. The obtained results make PANI-based memristive devices potentially widely applicable in neuromorphic systems.

The development of memristor-based spiking neuromorphic systems (NS) has been essentially driven by the hope to replicate the extremely high energy efficiency of biological systems. Spike-timing-dependent plasticity (STDP) mechanism is considered as one of the most promising learning rules for NS. STDP learning has been observed in different types of biological synapses in presence of neuromodulators, e.g. dopamine, and is believed to be an enabling phenomenon for important biological functions such as associative and reinforcement learning. However, the direct STDP window change under dopamine-like modulation has not been yet demonstrated in memristive synapses. In this study, we experimentally demonstrate a simple way for the STDP window shape modulation by introducing the coefficients controlling the neuron spike amplitudes. In such a way the STDP window shape could be modulated from a classical asymmetric shape to a bell-shaped, as well as to anti-STDP and to anti-bell-shaped. The experiments have been carried out with (Co_0.4Fe_0.4B_0.2)_x(LiNbO₃)_1−x nanocomposite-based memristors. Memristive characteristics of the nanocomposite structures with different metal content are also comprehensively studied. Obtained results give every hope for bio-inspired operation of the future large memristor-based NS with reinforcement learning ability.