Guided by brain-like ‘spiking’ computational frameworks, neuromorphic computing—brain-inspired computing for machine intelligence—promises to realize artificial intelligence while reducing the energy requirements of computing platforms. This interdisciplinary field began with the implementation of silicon circuits for biological neural routines, but has evolved to encompass the hardware implementation of algorithms with spike-based encoding and event-driven representations. Here we provide an overview of the developments in neuromorphic computing for both algorithms and hardware and highlight the fundamentals of learning and hardware frameworks. We discuss the main challenges and the future prospects of neuromorphic computing, with emphasis on algorithm–hardware codesign. The authors review the advantages and future prospects of neuromorphic computing, a multidisciplinary engineering concept for energy-efficient artificial intelligence with brain-inspired functionality.

Towards spike-based machine intelligence with neuromorphic computing.

Event cameras are bio-inspired sensors that differ from conventional frame cameras: Instead of capturing images at a fixed rate, they asynchronously measure per-pixel brightness changes, and output a stream of events that encode the time, location and sign of the brightness changes. Event cameras offer attractive properties compared to traditional cameras: high temporal resolution (in the order of is), very high dynamic range (140dB vs. 60dB), low power consumption, and high pixel bandwidth (on the order of kHz) resulting in reduced motion blur. Hence, event cameras have a large potential for robotics and computer vision in challenging scenarios for traditional cameras, such as low-latency, high speed, and high dynamic range. However, novel methods are required to process the unconventional output of these sensors in order to unlock their potential. This paper provides a comprehensive overview of the emerging field of event-based vision, with a focus on the applications and the algorithms developed to unlock the outstanding properties of event cameras. We present event cameras from their working principle, the actual sensors that are available and the tasks that they have been used for, from low-level vision (feature detection and tracking, optic flow, etc.) to high-level vision (reconstruction, segmentation, recognition). We also discuss the techniques developed to process events, including learning-based techniques, as well as specialized processors for these novel sensors, such as spiking neural networks. Additionally, we highlight the challenges that remain to be tackled and the opportunities that lie ahead in the search for a more efficient, bio-inspired way for machines to perceive and interact with the world.

Event-based Vision: A Survey

Review on spintronics: Principles and device applications

Neuromorphic computing could address the inherent limitations of conventional silicon technology in dedicated machine learning applications. Recent work on silicon-based asynchronous spiking neural networks and large crossbar arrays of two-terminal memristive devices has led to the development of promising neuromorphic systems. However, delivering a compact and efficient parallel computing technology that is capable of embedding artificial neural networks in hardware remains a significant challenge. Organic electronic materials offer an attractive option for such systems and could provide biocompatible and relatively inexpensive neuromorphic devices with low-energy switching and excellent tunability. Here, we review the development of organic neuromorphic devices. We consider different resistance-switching mechanisms, which typically rely on electrochemical doping or charge trapping, and report approaches that enhance state retention and conductance tuning. We also discuss the challenges the field faces in implementing low-power neuromorphic computing, such as device downscaling and improving device speed. Finally, we highlight early demonstrations of device integration into arrays, and consider future directions and potential applications of this technology.

Organic electronics for neuromorphic computing

There are two general approaches to developing artificial general intelligence (AGI)1: computer-science-oriented and neuroscience-oriented. Because of the fundamental differences in their formulations and coding schemes, these two approaches rely on distinct and incompatible platforms2–8, retarding the development of AGI. A general platform that could support the prevailing computer-science-based artificial neural networks as well as neuroscience-inspired models and algorithms is highly desirable. Here we present the Tianjic chip, which integrates the two approaches to provide a hybrid, synergistic platform. The Tianjic chip adopts a many-core architecture, reconfigurable building blocks and a streamlined dataflow with hybrid coding schemes, and can not only accommodate computer-science-based machine-learning algorithms, but also easily implement brain-inspired circuits and several coding schemes. Using just one chip, we demonstrate the simultaneous processing of versatile algorithms and models in an unmanned bicycle system, realizing real-time object detection, tracking, voice control, obstacle avoidance and balance control. Our study is expected to stimulate AGI development by paving the way to more generalized hardware platforms. The ‘Tianjic’ hybrid electronic chip combines neuroscience-oriented and computer-science-oriented approaches to artificial general intelligence, demonstrated by controlling an unmanned bicycle.

Towards artificial general intelligence with hybrid Tianjic chip architecture.

Loihi is a 60-mm2 chip fabricated in Intels 14-nm process that advances the state-of-the-art modeling of spiking neural networks in silicon. It integrates a wide range of novel features for the field, such as hierarchical connectivity, dendritic compartments, synaptic delays, and, most importantly, programmable synaptic learning rules. Running a spiking convolutional form of the Locally Competitive Algorithm, Loihi can solve LASSO optimization problems with over three orders of magnitude superior energy-delay-product compared to conventional solvers running on a CPU iso-process/voltage/area. This provides an unambiguous example of spike-based computation, outperforming all known conventional solutions.

Loihi: A Neuromorphic Manycore Processor with On-Chip Learning

A multiply-accumulate unit, or MAC, may achieve high throughput. The MAC need not use redundant hardware, such as multiple Wallace trees, or pipelining logic, yet may perform Wallace tree and carry look-ahead adder functions simultaneously for different operations.

Processing multiply-accumulate operations in a single cycle

A high-performance and low-power 32-bit multiply-accumulate unit (MAC) is described in this paper. The last mixed-length encoding scheme used in the MAC leverages the advantage of a 16-bit encoding scheme without adding extra delay to the faster four-stage Wallace tree of a 12-bit encoding scheme. With this new encoding scheme, one-cycle throughput for 16-bit /spl times/16-bit and 32-bit /spl times/16-bit MAC instructions was achieved at very high frequencies. To handle media streams more efficiently, the single-instruction-multiple-data (SIMD) and the multiply-with-implicit-accumulate (MIA) features were added. A mixture of static CMOS logic and complementary pass-gate logic (CPL) was used to achieve the high-speed and low-power goals. Several power-saving techniques were also implemented in this MAC.

A high-performance and low-power 32-bit multiply-accumulate unit with single-instruction-multiple-data (SIMD) feature

An embodiment of the present invention is a mixed length encoding unit. The mixed length may be a 12/16 bits (12/16-b) encoding algorithm within a multiply-accumulate (MAC). The mixed length encoding unit includes 16-b Booth encoder adapted to produce eight partial product vectors from sixteen bits of data. The 16-b Booth encoder is coupled to a four stage Wallace Tree. During a first cycle of the invention, a multiplex system directs the eight partial products and an accumulation vector to a four stage Wallace Tree. During subsequent cycles, the multiplex system directs six partial product vectors, an accumulation vector, one carry-feedback input vector, and one sum-feedback input vector to the four stage Wallace Tree.

Fast 16-B early termination implementation for 32-B multiply-accumulate unit

Graphics workloads are highly dynamic in nature, using multi-threaded SIMD execution units (EUs), fixed-function units, samplers, and media accelerators to provide ever-increasing amounts of graphics performance. These workloads are often limited by power and thermal constraints, requiring dynamic voltage/frequency scaling (DVFS) of the graphics processor (GPU). This coarse-grain DVFS, driven by a power-management IC (PMIC) setting a shared rail voltage (V IN ), incurs performance loss while waiting for PLL re-lock and slow-rail voltage transitions. In addition, it does not allow a performance-critical unit (e.g. an EU) to use on demand a higher V/F (e.g. for EU turbo) without an energy penalty for the rest of the GPU.

An energy-efficient graphics processor featuring fine-grain DVFS with integrated voltage regulators, execution-unit turbo, and retentive sleep in 14nm tri-gate CMOS

Yuyun Liao

Papers

Loihi: A Neuromorphic Manycore Processor with On-Chip Learning

Processing multiply-accumulate operations in a single cycle

A high-performance and low-power 32-bit multiply-accumulate unit with single-instruction-multiple-data (SIMD) feature

Fast 16-B early termination implementation for 32-B multiply-accumulate unit

An energy-efficient graphics processor featuring fine-grain DVFS with integrated voltage regulators, execution-unit turbo, and retentive sleep in 14nm tri-gate CMOS