Neuromorphic Hardware Architecture Using the Neural Engineering Framework for Pattern Recognition

Abstract

We present a hardware architecture that uses the neural engineering framework (NEF) to implement large-scale neural networks on field programmable gate arrays (FPGAs) for performing massively parallel real-time pattern recognition. NEF is a framework that is capable of synthesising large-scale cognitive systems from subnetworks and we have previously presented an FPGA implementation of the NEF that successfully performs nonlinear mathematical computations. That work was developed based on a compact digital neural core, which consists of 64 neurons that are instantiated by a single physical neuron using a time-multiplexing approach. We have now scaled this approach up to build a pattern recognition system by combining identical neural cores together. As a proof of concept, we have developed a handwritten digit recognition system using the MNIST database and achieved a recognition rate of 96.55%. The system is implemented on a state-of-the-art FPGA and can process 5.12 million digits per second. The architecture and hardware optimisations presented offer high-speed and resource-efficient means for performing high-speed, neuromorphic, and massively parallel pattern recognition and classification tasks.

Digital realization of the proposed linear model of the Hodgkin‐Huxley neuron

“…Recently, an active field of research on the design and implementation of analog and digital neural networks on the hardware platforms has attracted a lot of attention . In this regard, different approaches have proposed to design and construct spike base computing systems which imitate the nervous system . This article can be considered as a new approach to design the digital hardware platform of biophysical neurons model.…”

“…Therefore, in this paper, the attempt was made to simply the Hodgkin‐Huxley model by preserving dynamical characteristics of the biophysical model, which allows us to have an efficient and low‐cost hardware design on FPGA. Due to the high operating efficiency of the hardware realization of computational neuroscience, the hardware implementation of computational neuron models is one of the most popular methods to engage neural calculations on engineering issues in practical applications …”

“…The field‐programmable gate array (FPGA) in digital implementation benefits from low‐cost hardware implementation, flexibility, and faster design . Therefore, in recent years, this hardware platform has been very much considered in the area of neural computing . Most recently, the development of hardware description language (HDL) and synthesis tools make the configurable devices more powerful and appropriate in the efficient hardware design for neuromorphic architectures …”

“…Due to the high operating efficiency of the hardware realization of computational neuroscience, the hardware implementation of computational neuron models is one of the most popular methods to engage neural calculations on engineering issues in practical applications. [22][23][24] In fact, due to the capacity of very large scale integration (VLSI) design in the implementation of nonlinear models, it can be more pragmatic for hardware implementation of spiking neuronal networks on hardware platforms with the lower area and power consumption. 25,26 However, high cost and design time restrict the application of this approach.…”

Social Moments: A Perspective on Interaction for Social Robotics

“…As social moments are tied to social norms, detecting social moments will require the ability to predict typical behavior [e.g., motion from DVS, see Gibson et al (2014)] and highlight deviations. Using neuromorphic processing of visual inputs, studies have achieved categorization of objects in less than 160 ms (Wang et al, 2017) or triggered robot responses in 4 cycles of a periodic event (Wiles et al, 2010). In addition, the interpretation of social moments requires the integration of information across multiple modalities, as multiple modalities contribute to the generation of social meaning (Mondada, 2016).…”