Problems involving discrete Markov Chains are solved mathematically using matrix methods. Recently, several research groups have demonstrated that matrix-vector multiplication can be performed analytically in a single time step with an electronic circuit that incorporates an open-loop memristor crossbar that is effectively a resistive random-access memory. Ielmini and co-workers have taken this a step further by demonstrating that linear algebraic systems can also be solved in a single time step using similar hardware with feedback. These two approaches can both be applied to Markov chains, in the first case using matrix-vector multiplication to compute successive updates to a discrete Markov process and in the second directly calculating the stationary distribution by solving a constrained eigenvector problem. We present circuit models for open-loop and feedback configurations, and perform detailed analyses that include memristor programming errors, thermal noise sources and element nonidealities in realistic circuit simulations to determine both the precision and accuracy of the analog solutions. We provide mathematical tools to formally describe the trade-offs in the circuit model between power consumption and the magnitude of errors. We compare the two approaches by analyzing Markov chains that lead to two different types of matrices, essentially random and ill-conditioned, and observe that ill-conditioned matrices suffer from significantly larger errors. We compare our analog results to those from digital computations and find a significant power efficiency advantage for the crossbar approach for similar precision results.
18 January 2023 in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
Various centrality measures have been proposed to identify the influence of each node in a complex network. Among the most popular ranking metrics, spectral measures stand out from the crowd. They rely on the computation of the dominant eigenvector of suitable matrices related to the graph: EigenCentrality, PageRank, Hyperlink Induced Topic Search (HITS) and Stochastic Approach for Link-Structure Analysis (SALSA). The simplest algorithm used to solve this linear algebra computation is the Power Method. It consists of multiple Matrix-Vector Multiplications (MVMs) and a normalization step to avoid divergent behaviours. In this work, we present an analog circuit used to accelerate the Power Iteration algorithm including current-mode termination for the memristor crossbars and a normalization circuit. The normalization step together with the feedback loop of the complete circuit ensure stability and convergence of the dominant eigenvector. We implement a transistor level peripheral circuitry around the memristor crossbar and take non-idealities such as wire parasitics, source driver resistance and finite memristor precision into account. We compute the different spectral centralities to demonstrate the performance of the system. We compare our results to the ones coming from the conventional digital computers and observe significant energy savings while maintaining a competitive accuracy.
Artificial Intelligence (AI) is the domain of large resource intensive data centers that limit access to a small community of developers [1-3]. Neuromorphic hardware promises dramatically improved space and energy efficiency for AI, but is presently only capable of low-accuracy operations, like inferencing in neural networks [4-6]. Core computing tasks of signal processing, neural network training and natural language processing demand far higher computing resolution, beyond that of existing neuromorphic circuit elements [7-9]. Here we introduce an analogue molecular memristor based on a Ru-complex of an azo-aromatic ligand with 14-bit resolution. The supramolecular electronic structure of the film enables voltage-driven ionic rearrangement facilitating 16,520 distinct analogue levels. These levels can be linearly and symmetrically updated or written individually in one time-step, substantially simplifying the weight update procedure over existing neuromorphic platforms [4]. The circuit elements are unidirectional, facilitating a selector-less 64×64 crossbar-based dot-product engine that enables vector-matrix multiplication, including Fourier transform, in a single time-step. We achieved >73 dB signal-to-noise-ratio, a four-order of magnitude improvement over the state-of-the-art [10-12], while consuming 460× less energy than digital computers [13]. Accelerators leveraging these molecular crossbars could transform neuromorphic computing, extending it beyond niche applications and reinforcing the core of digital electronics from the cloud to the edge [14,15].
Traditional wireless receivers underperform in complex propagation environments and in the presence of nonlinear interference. Although data-driven artificial intelligence models offer a promising alternative, their implementation on conventional digital processing platforms is often limited by energy consumption and latency. Here, we present an analog wireless receiver that leverages the feature extraction and learning capabilities of neural networks for channel equalization, filtering, signal routing, and symbol detection. This modular architecture eliminates the traditional mixer for down-conversion and simplifies its implementation using analog in-memory computing hardware, allowing direct decoding of data from high-frequency analog signals. We demonstrate the adaptability, energy efficiency and processing speed of the proposed receiver through a proof-of-concept hardware implementation using memristor crossbar arrays on a system-on-a-chip platform. The fully hardware-implemented receiver achieves a bit error rate of 10 −6 in detecting randomly generated symbols from distorted analog signals across multiple modulation schemes and carrier frequencies, highlighting its potential for future wireless communication systems for bandwidth-intensive and latency-sensitive applications.