-
Neuromorphic Processing: A Unifying Tutorial
Authors:
Hamid Soleimani,
Emmanuel. M. Drakakis
Abstract:
All systolic or distributed neuromorphic architectures require power-efficient processing nodes. In this paper, a unifying tutorial is presented which implements multiple neuromorphic processing elements using a systematic analog approach including synapse, neuron and astrocyte models. It is shown that the proposed approach can successfully synthesize multidimensional dynamical systems into analog…
▽ More
All systolic or distributed neuromorphic architectures require power-efficient processing nodes. In this paper, a unifying tutorial is presented which implements multiple neuromorphic processing elements using a systematic analog approach including synapse, neuron and astrocyte models. It is shown that the proposed approach can successfully synthesize multidimensional dynamical systems into analog circuitry with minimum effort.
△ Less
Submitted 13 August, 2021;
originally announced August 2021.
-
A Generalized Strong-Inversion CMOS Circuitry for Neuromorphic Applications
Authors:
Hamid Soleimani,
Emmanuel. M. Drakakis
Abstract:
It has always been a challenge in the neuromorphic field to systematically translate biological models into analog electronic circuitry. In this paper, a generalized circuit design platform is introduced where biological models can be conveniently implemented using CMOS circuitry operating in strong-inversion. The application of the method is demonstrated by synthesizing a relatively complex two-d…
▽ More
It has always been a challenge in the neuromorphic field to systematically translate biological models into analog electronic circuitry. In this paper, a generalized circuit design platform is introduced where biological models can be conveniently implemented using CMOS circuitry operating in strong-inversion. The application of the method is demonstrated by synthesizing a relatively complex two-dimensional (2-D) nonlinear neuron model. The validity of our approach is verified by nominal simulated results with realistic process parameters from the commercially available AMS 0.35 um technology. The circuit simulation results exhibit regular spiking responses in good agreement with their mathematical counterpart.
△ Less
Submitted 6 August, 2021; v1 submitted 27 July, 2020;
originally announced July 2020.
-
A High GOPs/Slice Time Series Classifier for Portable and Embedded Biomedical Applications
Authors:
Hamid Soleimani,
Aliasghar,
Makhlooghpour,
Wilten Nicola,
Claudia Clopath,
Emmanuel. M. Drakakis
Abstract:
Nowadays a diverse range of physiological data can be captured continuously for various applications in particular wellbeing and healthcare. Such data require efficient methods for classification and analysis. Deep learning algorithms have shown remarkable potential regarding such analyses, however, the use of these algorithms on low-power wearable devices is challenged by resource constraints suc…
▽ More
Nowadays a diverse range of physiological data can be captured continuously for various applications in particular wellbeing and healthcare. Such data require efficient methods for classification and analysis. Deep learning algorithms have shown remarkable potential regarding such analyses, however, the use of these algorithms on low-power wearable devices is challenged by resource constraints such as area and power consumption. Most of the available on-chip deep learning processors contain complex and dense hardware architectures in order to achieve the highest possible throughput. Such a trend in hardware design may not be efficient in applications where on-node computation is required and the focus is more on the area and power efficiency as in the case of portable and embedded biomedical devices. This paper presents an efficient time-series classifier capable of automatically detecting effective features and classifying the input signals in real-time. In the proposed classifier, throughput is traded off with hardware complexity and cost using resource sharing techniques. A Convolutional Neural Network (CNN) is employed to extract input features and then a Long-Short-Term-Memory (LSTM) architecture with ternary weight precision classifies the input signals according to the extracted features. Hardware implementation on a Xilinx FPGA confirm that the proposed hardware can accurately classify multiple complex biomedical time series data with low area and power consumption and outperform all previously presented state-of-the-art records. Most notably, our classifier reaches 1.3$\times$ higher GOPs/Slice than similar state of the art FPGA-based accelerators.
△ Less
Submitted 1 November, 2018; v1 submitted 26 February, 2018;
originally announced February 2018.
-
Window functions and sigmoidal behaviour of memristive systems
Authors:
Panayiotis S. Georgiou,
Sophia N. Yaliraki,
Emmanuel M. Drakakis,
Mauricio Barahona
Abstract:
A common approach to model memristive systems is to include empirical window functions to describe edge effects and non-linearities in the change of the memristance. We demonstrate that under quite general conditions, each window function can be associated with a sigmoidal curve relating the normalised time-dependent memristance to the time integral of the input. Conversely, this explicit relation…
▽ More
A common approach to model memristive systems is to include empirical window functions to describe edge effects and non-linearities in the change of the memristance. We demonstrate that under quite general conditions, each window function can be associated with a sigmoidal curve relating the normalised time-dependent memristance to the time integral of the input. Conversely, this explicit relation allows us to derive window functions suitable for the mesoscopic modelling of memristive systems from a variety of well-known sigmoidals. Such sigmoidal curves are defined in terms of measured variables and can thus be extracted from input and output signals of a device and then transformed to its corresponding window. We also introduce a new generalised window function that allows the flexible modelling of asymmetric edge effects in a simple manner.
△ Less
Submitted 14 January, 2016;
originally announced May 2016.
-
Quantitative Measure of Hysteresis for Memristors Through Explicit Dynamics
Authors:
Panayiotis S. Georgiou,
Sophia N. Yaliraki,
Emmanuel M. Drakakis,
Mauricio Barahona
Abstract:
We introduce a mathematical framework for the analysis of the input-output dynamics of externally driven memristors. We show that, under general assumptions, their dynamics comply with a Bernoulli differential equation and hence can be nonlinearly transformed into a formally solvable linear equation. The Bernoulli formalism, which applies to both charge- and flux-controlled memristors when either…
▽ More
We introduce a mathematical framework for the analysis of the input-output dynamics of externally driven memristors. We show that, under general assumptions, their dynamics comply with a Bernoulli differential equation and hence can be nonlinearly transformed into a formally solvable linear equation. The Bernoulli formalism, which applies to both charge- and flux-controlled memristors when either current- or voltage-driven, can, in some cases, lead to expressions of the output of the device as an explicit function of the input. We apply our framework to obtain analytical solutions of the i-v characteristics of the recently proposed model of the Hewlett-Packard memristor under three different drives without the need for numerical simulations. Our explicit solutions allow us to identify a dimensionless lumped parameter that combines device-specific parameters with properties of the input drive. This parameter governs the memristive behavior of the device and, consequently, the amount of hysteresis in the i-v. We proceed further by defining formally a quantitative measure for the hysteresis of the device for which we obtain explicit formulas in terms of the aforementioned parameter and we discuss the applicability of the analysis for the design and analysis of memristor devices.
△ Less
Submitted 17 July, 2011; v1 submitted 30 October, 2010;
originally announced November 2010.
