A scalable approximate multiplier, called truncation- and rounding-based scalable approximate multiplier (TOSAM) is presented, which reduces the number of partial products by truncating each of the input operands based on their leading one-bit position. In the proposed design, multiplication is performed by shift, add, and small fixed-width multiplication operations resulting in large improvements in the energy consumption and area occupation compared to those of the exact multiplier. To improve the total accuracy, input operands of the multiplication part are rounded to the nearest odd number. Because input operands are truncated based on their leading one-bit positions, the accuracy becomes weakly dependent on the width of the input operands and the multiplier becomes scalable. Higher improvements in design parameters (e.g., area and energy consumption) can be achieved as the input operand widths increase. To evaluate the efficiency of the proposed approximate multiplier, its design parameters are compared with those of an exact multiplier and some other recently proposed approximate multipliers. Results reveal that the proposed approximate multiplier with a mean absolute relative error in the range of 11%-0.3% improves delay, area, and energy consumption up to 41%, 90%, and 98%, respectively, compared to those of the exact multiplier. It also outperforms other approximate multipliers in terms of speed, area, and energy consumption. The proposed approximate multiplier has an almost Gaussian error distribution with a near-zero mean value. We exploit it in the structure of a JPEG encoder, sharpening, and classification applications. The results indicate that the quality degradation of the output is negligible. In addition, we suggest an accuracy configurable TOSAM where the energy consumption of the multiplication operation can be adjusted based on the minimum required accuracy.
1 October 2017 in Computers and Electrical Engineering
In this article, the accuracy of inverter-based memristive neural networks (NNs) for function approximation applications is improved under the presence of process variations. The improvement is achieved by using a design approach, called INTERSTICE (Inverter-based Memristive Neural Networks Dis cretization for Function Approximation Applications), which discretizes the output values by employing a classifier. More precisely, in the INTERSTICE approach, the output range is divided into K subranges where each subrange is considered as a class. To train the classifier, the training samples are labeled where each label shows belonging to a specific class. To evaluate the efficacy of the design technique, some function approximation applications such as BlackScholes, FFT, K-means, and Sobel are considered. Compared to PHAX, a recently published inverter-based memristive NN, INTERSTICE provides lower mean squared error (MSE) values in the presence of memristor and transistor variations. More specifically, the improvements in the mean of MSE (μMSE) are in the range of 40%-80% when considering 10% variations in the memristor resistance and transistor parameters. In addition, for most of the benchmarks, INTERSTICE improves the μMSE values of the nominal case (the case where all circuit elements are ideal) compared to PHAX. As another advantage compared to the PHAX, in INTERSTICE, digital outputs can be generated based on the selected classes which eliminates the need for an analog-to-digital converter at the output port connected to the digital part of the system. Finally, achieving lower μMSE values using fewer memristors and consuming lower energy is also attainable with this design approach.
24 August 2021 in IEEE Transactions on Circuits and Systems I: Regular Papers
In this paper, the sensitivity of the neural network (NN) outputs to device parameter uncertainties (non-idealities) in inverter-based memristor ( IM ) crossbar neuromorphic circuits is mathematically modeled and verified using exhaustive circuit and system-level simulations. The NN sensitivity is obtained by modeling the sensitivity of the IM neuron output to the non-idealities of its circuit elements. The analysis reveals a higher sensitivity of the output voltage of the IM neuron to the non-idealities of the inverters compared to the conductance variation of the memristors. Among the inverter non-idealities, horizontal shift of the inverters voltage transfer characteristic (VTC) shows the highest impact on the output voltage of the neuron. To reduce the accuracy loss due to the variations, a training approach which includes a sensitivity term in the cost function of the training phase, is suggested. The achievable improvements through the said NN training approach are evaluated. In the evaluation, the California Housing, MNIST, and Fashion MNIST datasets are employed. The results show up to 50% reduction in the NN output variations in the presence of circuit elements’ non-idealities.
9 April 2021 in IEEE Transactions on Circuits and Systems II: Express Briefs
In this brief, we present a high accuracy training method for inverter-based memristive neural networks ( IM -NNs). The method, which relies on accurate modeling of the circuit element characteristics, is called LATIM (Loading-Aware offline Training method for Inverter-based Memristive NNs). In LATIM, an approximation method is proposed to estimate the effective load of the memristive crossbar (as the synapses) while two NNs are utilized to predict the voltage transfer characteristic (VTC) of the inverters (as the activation functions). Efficacy of the proposed method is compared with the recent offline training methods for IM -NNs, called PHAX and RIM. Simulation results reveal that LATIM can predict the output voltage of the IM -NNs, on average, by $14\times $ ( $6\times $ ) and $29\times $ ( $4\times $ ) smaller error for the MNIST and Fashion MNIST datasets, respectively, compared to those of PHAX (RIM) method. In addition, IM -NNs trained by LATIM consume, on average, 62% and 53% lower energy compared to PHAX and RIM methods due to proper sizing of the inverters.