Search

Current hardware approaches to biomimetic or neuromorphic artificial intelligence rely on elaborate transistor circuits to simulate biological functions. However, these can instead be more faithfully emulated by higher-order circuit elements that naturally express neuromorphic nonlinear dynamics.sup.1-4. Generating neuromorphic action potentials in a circuit element theoretically requires a minimum of third-order complexity (for example, three dynamical electrophysical processes).sup.5, but there have been few examples of second-order neuromorphic elements, and no previous demonstration of any isolated third-order element.sup.6-8. Using both experiments and modelling, here we show how multiple electrophysical processes--including Mott transition dynamics--form a nanoscale third-order circuit element. We demonstrate simple transistorless networks of third-order elements that perform Boolean operations and find analogue solutions to a computationally hard graph-partitioning problem. This work paves a way towards very compact and densely functional neuromorphic computing primitives, and energy-efficient validation of neuroscientific models. Electrophysical processes are used to create third-order nanoscale circuit elements, and these are used to realize a transistorless network that can perform Boolean operations and find solutions to a computationally hard graph-partitioning problem., Author(s): Suhas Kumar [sup.1] , R. Stanley Williams [sup.2] , Ziwen Wang [sup.3] Author Affiliations: (1) Hewlett Packard Labs, Palo Alto, USA (2) Texas A&M University, College Station, USA (3) [...]

Memristor-enabled neuromorphic computing systems provide a fast and energy-efficient approach to training neural networks.sup.1-4. However, convolutional neural networks (CNNs)--one of the most important models for image recognition.sup.5--have not yet been fully hardware-implemented using memristor crossbars, which are cross-point arrays with a memristor device at each intersection. Moreover, achieving software-comparable results is highly challenging owing to the poor yield, large variation and other non-ideal characteristics of devices.sup.6-9. Here we report the fabrication of high-yield, high-performance and uniform memristor crossbar arrays for the implementation of CNNs, which integrate eight 2,048-cell memristor arrays to improve parallel-computing efficiency. In addition, we propose an effective hybrid-training method to adapt to device imperfections and improve the overall system performance. We built a five-layer memristor-based CNN to perform MNIST.sup.10 image recognition, and achieved a high accuracy of more than 96 per cent. In addition to parallel convolutions using different kernels with shared inputs, replication of multiple identical kernels in memristor arrays was demonstrated for processing different inputs in parallel. The memristor-based CNN neuromorphic system has an energy efficiency more than two orders of magnitude greater than that of state-of-the-art graphics-processing units, and is shown to be scalable to larger networks, such as residual neural networks. Our results are expected to enable a viable memristor-based non-von Neumann hardware solution for deep neural networks and edge computing. A fully hardware-based memristor convolutional neural network using a hybrid training method achieves an energy efficiency more than two orders of magnitude greater than that of graphics-processing units., Author(s): Peng Yao [sup.1] , Huaqiang Wu [sup.1] [sup.2] , Bin Gao [sup.1] [sup.2] , Jianshi Tang [sup.1] [sup.2] , Qingtian Zhang [sup.1] , Wenqiang Zhang [sup.1] , J. Joshua [...]

The initial blank in the deep drawing process has a simple shape. After drawing, its perimeter shape becomes very complex. If the initial blank shape is designed in such a way that it is formed into the desired shape after the drawing process, not only does it reduces the time of trimming process but it also decreases the raw material needed substantially. In this paper, the genetically optimized neural network system (GONNS) is proposed as a tool to predict the initial blank shape for the desired final shape. Artificial neural networks (ANNs) represent the final blank shape after a training process and genetic algorithms find the optimum initial blank. The finite element method is employed for simulating the multilayer plate deep drawing process to provide training data for ANN. The GONNS results were verified through experiment in which the error was found to be about 0.2 mm. At last, variations of deformation force, thickness distribution, and thickness strain distribution were investigated using optimum blank. The results show 12% reduction in deformation force and more uniform thickness distribution as well as more consistent thickness strain distribution in the optimum blank shape. [DOI: 10.1115/1.4003121] Keywords: deep drawing, multilayer sheets, optimal blank, GONNS, FEM