Low-power, low-mismatch, highly-dense array of VLSI Mihalas-Niebur neurons

Hsu

2021 IEEE International Symposium on Circuits and Systems (ISCAS)

et al. 2021

The inner operations of the human brain as a biological processing system remain largely a mystery. Inspired by the function of the human brain and based on the analysis of simple neural network systems in other species, such as Drosophila, neuromorphic computing systems have attracted considerable interest. In cellular-level connectomics research, we can identify the characteristics of biological neural network, called population, which constitute not only recurrent fullyconnection in network, also an external-stimulus and selfconnection in each neuron. Relying on low data bandwidth of spike transmission in network and input data, Spiking Neural Networks exhibit low-latency and low-power design. In this study, we proposed a configurable population-based digital spiking neuromorphic processor in 180nm process technology with two configurable hierarchy populations. Also, these neurons in the processor can be configured as novel models, integer quadratic integrate-and-fire neuron models, which contain an unsigned 8-bit membrane potential value. The processor can implement intelligent decision making for avoidance in real-time. Moreover, the proposed approach enables the developments of biomimetic neuromorphic system and various low-power, and low-latency inference processing applications with normalized energy efficiency of 13.2 pJ/SOP.

Section: Specification Measurement and Resultsmentioning

confidence: 99%

POPPINS: A Population-Based Digital Spiking Neuromorphic Processor with Integer Quadratic Integrate-and-Fire Neurons

Hsu

2021 IEEE International Symposium on Circuits and Systems (ISCAS)

et al. 2021

“…We can envision the complete system integrated using the three separate blocks to demonstrate the real time applications. We already have the in-house built cochlea model on FPGA [45], which can be interfaced with the 2D-MoS 2 based memory array to implement the STDP learning capability, followed by a linear classifier system [49][50][51][52][53][54][55]. The current limitation is to integrate the memristive devices with the CMOS process on a common substrate.…”

Section: Discussionmentioning

confidence: 99%

Demonstration of intrinsic STDP learning capability in all-2D multi-state MoS₂ memory and its application in modelling neuromorphic speech recognition

Paul¹,

Mukundan²,

Tiwari³

et al. 2021

2D Mater.

The human brain can be characterized by its large number of adaptive synapses, connecting billions of neurons capable of both learning and perceiving the environment. Neuromorphic computing, based on brain-inspired principles, is a promising technology, to build low-power, distributed, fault-tolerant intelligent systems mainly for perception tasks. Here, we demonstrate the intrinsic capability of floating gate (FG) MoS2 device (MoS2 FG-FET) to model the spike time dependent plasticity (STDP) learning rule that is based on the transient response of the MoS2 channel to spikes applied to the source and gate leads. We implemented the STDP learning protocol in a neuromorphic speech recognition system (NSRS), inspired by the human auditory pathway, for various auditory recognition tasks. Our proposed NSRS consists of a cochlea model, an unsupervised feature learning stage, and a simple linear classifier. The unsupervised learning stage uses the biologically plausible STDP learning in novel two-dimensional MoS2 FG-FET memory which circumvents the requirement of any other learning circuitry. Demonstration of STDP modelling in two-dimensional (2D) MoS2 is an important step towards incorporating 2D architectures for reduced device footprints in neuromorphic learning circuits.

“…All neurons presented in this figure are standalone (i.e. non time-multiplexed), except in [137] for which only the update logic area is reported, and in [133] for which contributions from shared soma and threshold adaptation circuits are excluded. The designs from [56], [58], [96], [130] emulate an adaptiveexponential neuron model and are thus believed to reach the 20 Izhikevich behaviors [131], though not demonstrated.…”

Section: ) Neurons (Soma)mentioning

confidence: 99%

“…Niebur model in [133] and [134] were demonstrated to exhibit 9 and 15 out of the 20 Izhikevich behaviors, respectively, although with relatively high areas due to the older technology node used. The Morris-Lecar model is also explored in [129] and is believed to reach 13 out of the 20 Izhikevich behaviors [115].…”

Section: ) Neurons (Soma)mentioning

confidence: 99%

Bottom-Up and Top-Down Neural Processing Systems Design: Neuromorphic Intelligence as the Convergence of Natural and Artificial Intelligence

Frenkel¹,

Bol²,

Indiveri³

2021

Preprint

While Moore's law has driven exponential computing power expectations, its nearing end calls for new avenues for improving the overall system performance. One of these avenues is the exploration of new alternative brain-inspired computing architectures that promise to achieve the flexibility and computational efficiency of biological neural processing systems. Within this context, neuromorphic intelligence represents a paradigm shift in computing based on the implementation of spiking neural network architectures tightly co-locating processing and memory. In this paper, we provide a comprehensive overview of the field, highlighting the different levels of granularity present in existing silicon implementations, comparing approaches that aim at replicating natural intelligence (bottom-up) versus those that aim at solving practical artificial intelligence applications (top-down), and assessing the benefits of the different circuit design styles used to achieve these goals. First, we present the analog, mixed-signal and digital circuit design styles, identifying the boundary between processing and memory through time multiplexing, in-memory computation and novel devices. Next, we highlight the key tradeoffs for each of the bottom-up and top-down approaches, survey their silicon implementations, and carry out detailed comparative analyses to extract design guidelines. Finally, we identify both necessary synergies and missing elements required to achieve a competitive advantage for neuromorphic edge computing over conventional machine-learning accelerators, and outline the key elements for a framework toward neuromorphic intelligence.