16 August 2022 in IEEE Embedded Systems Letters
Approximate computing is an evolving paradigm that aims to improve the power, speed, and area in neural network applications that can tolerate errors up to a specific limit. This letter proposes a new multiplier architecture based on the algorithm that adapts the approximate compressor from the existing and proposed compressors’ set to reduce error in the respective partial product columns. Further, the error due to the approximation in the proposed multiplier is corrected using a simple error-correcting module. Results prove that the power and power–delay product (PDP) of an 8-bit multiplier is improved by up to 39.9% and 43.6% compared with the exact multiplier and 27.5% and 23.9% compared to similar previous designs. The proposed multiplier is validated using image processing and neural network applications to prove its efficacy.
16 December 2020 in IEEE Embedded Systems Letters
Approximate computing is a promising method for designing power-efficient computing systems. Many image and compression algorithms are inherently error-tolerant and can allow errors up to a specific limit. In such algorithms, savings in power can be achieved by approximating the data path units, such as a multiplier. This letter presents a novel decoder logic-based multiplier design with the intent to reduce the partial products generated. Thus, leading to a reduction in the hardware complexity and power consumption while maintaining a low error rate. Our proposed design in an 8-bit format which achieves 40.96% and 22.30% power reduction compared to the accurate and approximate multipliers. Comprehensive simulations are carried out on image sharpening and compression algorithms to prove that the proposed design obtains a better quality-effort tradeoff than the existing multipliers.
20 July 2021 in IEEE Access
The high computational complexity, memory footprints, and energy requirements of machine learning models, such as Artificial Neural Networks (ANNs), hinder their deployment on resource-constrained embedded systems. Most state-of-the-art works have considered this problem by proposing various low bit-width data representation schemes and optimized arithmetic operators’ implementations. To further elevate the implementation gains offered by these individual techniques, there is a need to cross-examine and combine these techniques’ unique features. This paper presents ExPAN(N)D, a framework to analyze and ingather the efficacy of the Posit number representation scheme and the efficiency of fixed-point arithmetic implementations for ANNs. The Posit scheme offers a better dynamic range and higher precision for various applications than IEEE 754 single-precision floating-point format. However, due to the dynamic nature of the various fields of the Posit scheme, the corresponding arithmetic circuits have higher critical path delay and resource requirements than the single-precision-based arithmetic units. Towards this end, we propose a novel Posit to fixed-point converter for enabling high-performance and energy-efficient hardware implementations for ANNs with minimal drop in the output accuracy. We also propose a modified Posit-based representation to store the trained parameters of a network. With the proposed Posit to fixed-point converter-based designs, we provide multiple design points with varying accuracy-performance trade-offs for an ANN. For instance, compared to the lowest power dissipating Posit-only accelerator design, one of our proposed designs results in 80% and 48% reduction in power dissipation and LUT utilization respectively, with marginal increase in classification error for Imagenet dataset classification using VGG-16.
20 December 2019 in IEEE Journal of Photovoltaics
Exploring the underwater solar energy by solar photovoltaic (PV) cells leads to a huge advantage by utilizing the humongous space of water covered by the earth's surface. Even though the amount of solar radiation decreases with the depth of the water, water provides sustainable cooling and cleaning for solar PV cells underwater. There are many challenges and constraints to develop solar PV cells underwater because they are mostly calibrated and amenable to space, dryland, terrestrial, etc., and the solar spectrum is prone to get narrower with the depth of the water. The implementation of solar PV cells underwater is pliable in various commercial and defense applications, such as sensors, water monitoring systems, autonomous vehicles, underwater gliders, etc. In this article, first, a mathematical model has been developed for the solar cell spectrum to incorporate the changes in the solar irradiance with the depth of the water. Furthermore, an experimental setup was designed and implemented to mimic an underwater environment. The performance of the polycrystalline encapsulated solar cell was studied based on the different types of water and the depth of the solar cell underwater. This article manifests that there is a sufficient amount of underwater solar power that can be utilized using PV cells to operate various devices and systems.
1 September 2023 in Flexible and Printed Electronics
While laser-induced graphitic carbon (LIGC) has been used to fabricate cost-effective conductive carbon on flexible substrates for applications such as sensors and energy storage devices, predicting the resistance of the component fabricated via LIGC remains challenging. In this study, a two-step machine learning-based modeling framework is developed to predict the sheet resistance of the materials fabricated using LIGC. The two-step modeling framework consists of classification and regression. First, random forest (RF) is used to classify successful and failed trials. Second, extreme gradient boosting (XGBoost), RF, support vector machine with radial basis function, multivariate adaptive spline regression, and multilayer perceptron are used to predict the sheet resistance in each successful trial. In addition, an analysis of the change in sheet resistance with respect to laser energy per unit area is conducted to remove data points with high sheet resistance. XGBoost is also used to determine the importance of each process parameter. We demonstrate the modeling framework on datasets collected from experiments where LIGC lines (1D) and LIGC squares (2D) are engraved. For the 1D dataset, the RF classification model achieves a 95% accuracy. For both 1D and 2D datasets, a comparative study shows that XGBoost outperforms other algorithms. XGBoost predicts the sheet resistance of the LIGC lines and squares with a MAPE of 7.08% and 8.75%, respectively. XGBoost also identifies laser resolution as the most significant parameter. Moreover, experimental results show that models built on the dataset merging the 1D and 2D datasets result in lower prediction accuracy than those built on the 1D and 2D datasets separately. The modeling framework allows one to determine the sheet resistance of LIGC with varying laser processing conditions without conducting expensive and time-consuming experiments.