Locality of interactions for planar memristive circuits

An analysis of the literature shows that there are two types of non-memristive models that have been widely used in the modeling of so-called "memristive" neural networks. Here, we demonstrate that such models have nothing in common with the concept of memristive elements: they describe either non-linear resistors or certain bi-state systems, which all are devices without memory. Therefore, the results presented in a significant number of publications are at least questionabl...

The memristor is a device whose resistance changes depending on the polarity and magnitude of a voltage applied to the device's terminals. We design a minimalistic model of a regular network of memristors using structurally-dynamic cellular automata. Each cell gets info about states of its closest neighbours via incoming links. A link can be one 'conductive' or 'non-conductive' states. States of every link are updated depending on states of cells the link connects. Every cell...

We describe a hybrid analog-digital computing approach to solve important combinatorial optimization problems that leverages memristors (two-terminal nonvolatile memories). While previous memristor accelerators have had to minimize analog noise effects, we show that our optimization solver harnesses such noise as a computing resource. Here we describe a memristor-Hopfield Neural Network (mem-HNN) with massively parallel operations performed in a dense crossbar array. We provi...

Despite all the progress of semiconductor integrated circuit technology, the extreme complexity of the human cerebral cortex makes the hardware implementation of neuromorphic networks with a comparable number of devices exceptionally challenging. One of the most prospective candidates to provide comparable complexity, while operating much faster and with manageable power dissipation, are so-called CrossNets based on hybrid CMOS/memristor circuits. In these circuits, the usual...

It is observed that the inductive and capacitive features of the memristor reflect (and are a quintessence of) such features of any resistor. The very presence of the voltage and current state variables, associated by their electrodynamics sense with electrical and magnetic fields, in the resistive characteristic v = f(i), forces any resister to accumulate some magnetic and electrostatic fields and energies around itself, i.e. L and C elements are always present. From the cir...

We suggest and experimentally demonstrate a chaotic memory resistor (memristor). The core of our approach is to use a resistive system whose equations of motion for its internal state variables are similar to those describing a particle in a multi-well potential. Using a memristor emulator, the chaotic memristor is realized and its chaotic properties are measured. A Poincar\'{e} plot showing chaos is presented for a simple nonautonomous circuit involving only a voltage source...

In this paper we revisit the memristor concept within circuit theory. We start from the definition of the basic circuit elements, then we introduce the original formulation of the memristor concept and summarize some of the controversies on its nature. We also point out the ambiguities resulting from a non rigorous usage of the flux linkage concept. After concluding that the memristor is not a fourth basic circuit element, prompted by recent claims in the memristor literature...

In-memory computing with crosspoint resistive memory arrays has been shown to accelerate data-centric computations such as the training and inference of deep neural networks, thanks to the high parallelism endowed by physical rules in the electrical circuits. By connecting crosspoint arrays with negative feedback amplifiers, it is possible to solve linear algebraic problems such as linear systems and matrix eigenvectors in just one step. Based on the theory of feedback circui...

Memristive systems, namely resistive systems with memory, are attracting considerable attention due to their ubiquity in several phenomena and technological applications. Here, we show that even the simplest one-dimensional network formed by the most common memristive elements with voltage threshold bears non-trivial physical properties. In particular, by taking into account the single element variability we find i) dynamical acceleration and slowing down of the total resista...

Recent breakthroughs in recurrent deep neural networks with long short-term memory (LSTM) units has led to major advances in artificial intelligence. State-of-the-art LSTM models with significantly increased complexity and a large number of parameters, however, have a bottleneck in computing power resulting from limited memory capacity and data communication bandwidth. Here we demonstrate experimentally that LSTM can be implemented with a memristor crossbar, which has a small...