I have developed "tennis elbow" from lugging this book around the past four weeks, but it is worth the pain, the effort, and the aspirin. It is also worth the (relatively speaking) bargain price. Including appendixes, this book contains 894 pages of text. The entire panorama of the neural sciences is surveyed and examined, and it is comprehensive in its scope, from genomes to social behaviors. The editors explicitly state that the book is designed as "an introductory text for students of biology, behavior, and medicine," but it is hard to imagine any audience, interested in any fragment of neuroscience at any level of sophistication, that would not enjoy this book. The editors have done a masterful job of weaving together the biologic, the behavioral, and the clinical sciences into a single tapestry in which everyone from the molecular biologist to the practicing psychiatrist can find and appreciate his or

Principles of Neural Science

The organization of behavior: A neuropsychological theory

Memristive devices are electrical resistance switches that can retain a state of internal resistance based on the history of applied voltage and current. These devices can store and process information, and offer several key performance characteristics that exceed conventional integrated circuit technology. An important class of memristive devices are two-terminal resistance switches based on ionic motion, which are built from a simple conductor/insulator/conductor thin-film stack. These devices were originally conceived in the late 1960s and recent progress has led to fast, low-energy, high-endurance devices that can be scaled down to less than 10 nm and stacked in three dimensions. However, the underlying device mechanisms remain unclear, which is a significant barrier to their widespread application. Here, we review recent progress in the development and understanding of memristive devices. We also examine the performance requirements for computing with memristive devices and detail how the outstanding challenges could be met.

Memristive devices for computing

Resistance switching in metal oxides could form the basis for next-generation non-volatile memory. It has been argued that the current in the high-conductivity state of several technologically relevant oxide materials flows through localized filaments, but these filaments have been characterized only indirectly, limiting our understanding of the switching mechanism. Here, we use high-resolution transmission electron microscopy to probe directly the nanofilaments in a Pt/TiO2/Pt system during resistive switching. In situ current–voltage and low-temperature (∼130 K) conductivity measurements confirm that switching occurs by the formation and disruption of TinO2n−1 (or so-called Magneli phase) filaments. Knowledge of the composition, structure and dimensions of these filaments will provide a foundation for unravelling the full mechanism of resistance switching in oxide thin films, and help guide research into the stability and scalability of such films for applications. Nanoscale filaments with a Magneli structure are shown to be responsible for resistance switching in thin films of TiO2, and the properties of the filaments are directly observed during the switching process.

Atomic structure of conducting nanofilaments in TiO2 resistive switching memory

We developed two-step solution-phase reactions to form hybrid materials of Mn3O4 nanoparticles on reduced graphene oxide (RGO) sheets for lithium ion battery applications. Mn3O4 nanoparticles grown selectively on RGO sheets over free particle growth in solution allowed for the electrically insulating Mn3O4 nanoparticles wired up to a current collector through the underlying conducting graphene network. The Mn3O4 nanoparticles formed on RGO show a high specific capacity up to ~900mAh/g near its theoretical capacity with good rate capability and cycling stability, owing to the intimate interactions between the graphene substrates and the Mn3O4 nanoparticles grown atop. The Mn3O4/RGO hybrid could be a promising candidate material for high-capacity, low-cost, and environmentally friendly anode for lithium ion batteries. Our growth-on-graphene approach should offer a new technique for design and synthesis of battery electrodes based on highly insulating materials.

Mn3O4-Graphene Hybrid as a High Capacity Anode Material for Lithium Ion Batteries

The niobium oxide based nonlinear devices were fabricated and analyzed. The electrical tests present good control over device-to-device variation and stable endurance more than 106. The conduction mechanism was comprehensively discussed. The low-voltage region relates to SCLC, while Schottky and F-N tunneling are responsible for negative and positive high-voltage region respectively. Consequently, the volatile switching may originate from the capture and release of electrons by traps in SCLC mechanism, and this relaxation phenomenon is evaluated quantitatively.

Investigation of NbO$_{\boldsymbol~x}$-based volatile switching device with self-rectifying characteristics

The RRAM-based accelerators enable fast and energy-efficient inference for neural networks. However, there are some requirements to deploy neural networks on RRAM-based accelerators, which are not considered in existing neural networks. (1) Because the noise problem and analog-digital converters/digital-analog converters (ADC/DAC) affect the prediction accuracy, they should be modeled in networks. (2) Because the weights are mapped to the RRAM cells, they should be quantized, and the number of weights is limited by the number of RRAM cells in the accelerator. These requirements motivate us to customize the hardware-friendly network for the RRAM-based accelerator. We take the idea of network architecture search (NAS) to design networks with high prediction accuracy that meet the requirements. We propose a framework called NAS4RRAM to search for the optimal network on the given RRAM-based accelerator. The experiments demonstrate that NAS4RRAM can apply to different RRAM-based accelerators with different scales. The performance of searched networks outperforms the manually designed ResNet.

NAS4RRAM: neural network architecture search for inference on RRAM-based accelerators

Memristor-based logic design gives an alternative solution to improve the energy efficiency of computing systems, benefiting from combining the memory with computing units. Inspired by this thought, previous work has demonstrated various memristor-based logic families with different attributes and computation patterns. Besides, some logic synthesis tools are designed for specific memristive logic implementations. However, the poor universality and the neglect of realistic constraints in memory largely restrict the utility of these logic synthesis tools. In this paper, we propose a general logic synthesis framework for memristor-based logic design, containing a universal abstract description method for memristive logic, a mapping rules generator, and a synthesis and mapping flow. The proposed logic synthesis framework is suitable for various types of existing memristor-based logic families and takes the memory status into consideration. It is also possible to handle future memristive devices and logic families by providing the universal abstraction interface. Furthermore, we also design a circuit-partitioning-based synthesis acceleration strategy to tackle with the long synthesis time problem. Experimental results show that, our framework can generate mapping results under the restriction of limited resource, while the existing synthesis tools may fail under the same restriction, and achieve comparable synthesis results with the same resource as the existing synthesis tools, which is enough for computation and storage. And the proposed acceleration scheme can achieve ∼ 1000× speedup compared with the initial one.

A General Logic Synthesis Framework for Memristor-based Logic Design

Simultaneous electrodeposition of metals facilitates the realization of memrisitive devices with high yield and improved reliability.

Memristors with alloyed electrodes

The recent advances in optic neuromorphic devices have led to a subsequent rise in use for construction of energy-efficient artificial vision systems. The widespread use can be attributed to their ability to capture, store, and process visual information from the environment. The primary limitations of existing optic neuromorphic devices include nonlinear weight updates, cross-talk issues, and silicon process incompatibility. In this study, we experimentally demonstrate a highly-linear, light-tunable, cross-talk-free, and silicon-compatible one-phototransistor-one-memristor (1PT1R) optic memristor for the implementation of optic artificial neural networks (OANN). For optic image recognition in the experiment an OANN was constructed using a 16×3 1PT1R memristor array, and it was trained on an online platform. The model yielded an accuracy of 99.3% after only 10 training epochs. The 1PT1R memristor, which shows good performance, demonstrated its ability as an excellent hardware solution for highly efficient optic neuromorphic and edge computing. This article is protected by copyright. All rights reserved.

Yuchao Yang

Papers

Investigation of NbO$_{\boldsymbol~x}$-based volatile switching device with self-rectifying characteristics

NAS4RRAM: neural network architecture search for inference on RRAM-based accelerators

A General Logic Synthesis Framework for Memristor-based Logic Design

Memristors with alloyed electrodes

One-phototransistor-one-memristor Array with High-linearity Light-tunable Weight for Optic Neuromorphic Computing.