12 August 2020 in Proceedings of the IEEE
FSFrancisco Javier Hernandez Santiago
Francisco Javier Hernandez Santiago
Approximate computing has emerged as a new paradigm for high-performance and energy-efficient design of circuits and systems. For the many approximate arithmetic circuits proposed, it has become critical to understand a design or approximation technique for a specific application to improve performance and energy efficiency with a minimal loss in accuracy. This article aims to provide a comprehensive survey and a comparative evaluation of recently developed approximate arithmetic circuits under different design constraints. Specifically, approximate adders, multipliers, and dividers are synthesized and characterized under optimizations for performance and area. The error and circuit characteristics are then generalized for different classes of designs. The applications of these circuits in image processing and deep neural networks indicate that the circuits with lower error rates or error biases perform better in simple computations, such as the sum of products, whereas more complex accumulative computations that involve multiple matrix multiplications and convolutions are vulnerable to single-sided errors that lead to a large error bias in the computed result. Such complex computations are more sensitive to errors in addition than those in multiplication, so a larger approximation can be tolerated in multipliers than in adders. The use of approximate arithmetic circuits can improve the quality of image processing and deep learning in addition to the benefits in performance and power consumption for these applications.
2 May 2018 in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
Approximate computing has been considered to improve the accuracy-performance trade-off in error-tolerant applications. For many of these applications, multiplication is a key arithmetic operation. Given that approximate compressors are a key element in the design of power-efficient approximate multipliers, we first propose an initial approximate 4:2 compressor that introduces a rather large error to the output. However, the number of faulty rows in the compressor’s truth table is significantly reduced by encoding its inputs using generate and propagate signals. Based on this improved compressor, two 4×4 multipliers are designed with different accuracies and then are used as building blocks for scaling up to 16×16 and 32×32 multipliers. According to the mean relative error distance (MRED), the most accurate of the proposed 16×16 unsigned designs has a 44% smaller power-delay product (PDP) compared to other designs with comparable accuracy. The radix-4 signed Booth multiplier constructed using the proposed compressor achieves a 52% reduction in the PDP-MRED product compared to other approximate Booth multipliers with comparable accuracy. The proposed multipliers outperform other approximate designs in image sharpening and joint photographic experts group (JPEG) applications by achieving higher quality outputs with lower power consumptions. For the first time, we show the applicability and practicality of approximate multipliers in multiple-input multiple-output (MIMO) antenna communication systems with error control coding.
5 August 2020 in IEEE Transactions on Neural Networks and Learning Systems
Neural networks (NNs) are effective machine learning models that require significant hardware and energy consumption in their computing process. To implement NNs, stochastic computing (SC) has been proposed to achieve a tradeoff between hardware efficiency and computing performance. In an SC NN, hardware requirements and power consumption are significantly reduced by moderately sacrificing the inference accuracy and computation speed. With recent developments in SC techniques, however, the performance of SC NNs has substantially been improved, making it comparable with conventional binary designs yet by utilizing less hardware. In this article, we begin with the design of a basic SC neuron and then survey different types of SC NNs, including multilayer perceptrons, deep belief networks, convolutional NNs, and recurrent NNs. Recent progress in SC designs that further improve the hardware efficiency and performance of NNs is subsequently discussed. The generality and versatility of SC NNs are illustrated for both the training and inference processes. Finally, the advantages and challenges of SC NNs are discussed with respect to binary counterparts.
8 October 2019 in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
Improving the accuracy of a neural network (NN) usually requires using larger hardware that consumes more energy. However, the error tolerance of NNs and their applications allow approximate computing techniques to be applied to reduce implementation costs. Given that multiplication is the most resource-intensive and power-hungry operation in NNs, more economical approximate multipliers (AMs) can significantly reduce hardware costs. In this article, we show that using AMs can also improve the NN accuracy by introducing noise. We consider two categories of AMs: 1) deliberately designed and 2) Cartesian genetic programing (CGP)-based AMs. The exact multipliers in two representative NNs, a multilayer perceptron (MLP) and a convolutional NN (CNN), are replaced with approximate designs to evaluate their effect on the classification accuracy of the Mixed National Institute of Standards and Technology (MNIST) and Street View House Numbers (SVHN) data sets, respectively. Interestingly, up to 0.63% improvement in the classification accuracy is achieved with reductions of 71.45% and 61.55% in the energy consumption and area, respectively. Finally, the features in an AM are identified that tend to make one design outperform others with respect to NN accuracy. Those features are then used to train a predictor that indicates how well an AM is likely to work in an NN.
1 April 2021 in IEEE Transactions on Computers
Multiplication is the most resource-hungry operation in neural networks (NNs). Logarithmic multipliers (LMs) simplify multiplication to shift and addition operations and thus reduce the energy consumption. Since implementing the logarithm in a compact circuit often introduces approximation, some accuracy loss is inevitable in LMs. However, this loss is tolerated by the inherent error tolerance of NNs and their associated applications. This article proposes an improved logarithmic multiplier (ILM) that, unlike existing designs, rounds both inputs to their nearest powers of two by using a proposed nearest-one detector (NOD) circuit. Considering that the output of the NOD uses a one-hot representation, some entries in the truth table of a conventional adder cannot occur. Hence, a compact adder is designed for the reduced truth table. The 8×8 ILM achieves up to 17.48% saving in power consumption compared to a recent LM in the literature while being almost 15% more accurate. Moreover, the evaluation of the ILM for two benchmark NN workloads shows up to 21.85% reduction in energy consumption compared to the NNs implemented with other LMs. Interestingly, using the ILM increases the classification accuracy of the considered NNs by up to 1.4% compared to a NN implementation that uses exact multipliers.
29 March 2018 in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
Stochastic computing (SC) often requires long stochastic sequences and, thus, a long latency to achieve accurate computation. The long latency leads to an inferior performance and low energy efficiency compared with most conventional binary designs. In this paper, a type of low-discrepancy sequences, the Sobol sequence, is considered for use in SC. Compared to the use of pseudorandom sequences generated by linear feedback shift registers (LFSRs), the use of Sobol sequences improves the accuracy of stochastic computation with a reduced sequence length. The inherent feature in Sobol sequence generators enables the parallel implementation of random number generators with an improved performance and hardware efficiency. In particular, the underlying theory is formulated and circuit design is proposed for an arbitrary level of parallelization in a power of 2. In addition, different strategies are implemented for parallelizing combinational and sequential stochastic circuits. The hardware efficiency of the parallel stochastic circuits is measured by energy per operation (EPO), throughput per area (TPA), and runtime. At a similar accuracy, the 8× parallel stochastic circuits using Sobol sequences consume approximately 1% of the EPO of the conventional LFSR-based nonparallelized circuits. Meanwhile, an average of 70 (up to 89) times improvements in TPA and less than 1% runtime are achieved. A sorting network is implemented for a median filter (MF) as an application. For a similar image processing quality, a higher energy efficiency is obtained for an 8× parallelized stochastic MF compared with its binary counterpart.