Design and Analysis of Energy-Efficient Dynamic Range Approximate Logarithmic Multipliers for Machine Learning

Yin, Peipei; Wang, Chenghua; Waris, Haroon; Liu, Weiqiang; Han, Yinhe; Lombardi, Fabrizio

doi:10.1109/tsusc.2020.3004980

Cited by 43 publications

(20 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In general, the LEE corresponds to the successive accumulation of logarithms of a squared signal. To do so, a multiplier for obtaining the squared signal is used, while the logarithm is estimated using Mitchell’s algorithm [ 68 ] and logarithm properties. The -Log_2 block initiates the process through the signal STR and the signal that indicates the end of the process is RDY.…”

Section: Experiments and Resultsmentioning

confidence: 99%

Short-Circuited Turn Fault Diagnosis in Transformers by Using Vibration Signals, Statistical Time Features, and Support Vector Machines on FPGA

Huerta-Rosales

Granados-Lieberman

García-Pérez

et al. 2021

Sensors

View full text Add to dashboard Cite

One of the most critical devices in an electrical system is the transformer. It is continuously under different electrical and mechanical stresses that can produce failures in its components and other electrical network devices. The short-circuited turns (SCTs) are a common winding failure. This type of fault has been widely studied in literature employing the vibration signals produced in the transformer. Although promising results have been obtained, it is not a trivial task if different severity levels and a common high-level noise are considered. This paper presents a methodology based on statistical time features (STFs) and support vector machines (SVM) to diagnose a transformer under several SCTs conditions. As STFs, 19 indicators from the transformer vibration signals are computed; then, the most discriminant features are selected using the Fisher score analysis, and the linear discriminant analysis is used for dimension reduction. Finally, a support vector machine classifier is employed to carry out the diagnosis in an automatic way. Once the methodology has been developed, it is implemented on a field-programmable gate array (FPGA) to provide a system-on-a-chip solution. A modified transformer capable of emulating different SCTs severities is employed to validate and test the methodology and its FPGA implementation. Results demonstrate the effectiveness of the proposal for diagnosing the transformer condition as an accuracy of 96.82% is obtained.

show abstract

Section: Experiments and Resultsmentioning

confidence: 99%

Short-Circuited Turn Fault Diagnosis in Transformers by Using Vibration Signals, Statistical Time Features, and Support Vector Machines on FPGA

Huerta-Rosales

Granados-Lieberman

García-Pérez

et al. 2021

Sensors

View full text Add to dashboard Cite

show abstract

“…We analyzed and compared the hardware performance of the 16-bit signed fixed-point GEMM and AGEMM units in terms of the power, area, delay, and power delay product (PDP). We compared the GEMM unit using the exact radix-4 multipliers to the AGEMM units employing the logarithmic multipliers DR-ALM5 [ 57 ] and TL16-8/4 [ 61 ], the nonlogarithmic multiplier RAD1024 [ 55 ], and the hybrid HRALM3 multiplier [ 62 ].…”

Section: An Approximate General Matrix Multiply Unitmentioning

confidence: 99%

“…Multiplication is a very common, but expensive operation, with exact multipliers being large circuits that consume a significant amount of energy. Various approximate multipliers have been proposed in recent years [ 52 , 53 , 54 , 55 , 56 , 57 , 58 , 59 , 60 , 61 , 62 ]. Many studies reported that approximate multipliers behave well in neural network processing [ 56 , 59 , 60 , 61 , 63 , 64 , 65 ].…”

Section: Introductionmentioning

confidence: 99%

“…In the approximate multiplier design, we balanced accuracy and energy efficiency to suit the application’s needs. Various approximate multipliers have been proposed in recent years [ 54 , 55 , 56 , 57 , 58 , 59 , 60 , 61 , 62 ]. Approximate multipliers follow one of three design strategies: the approximate logarithmic design, the approximate nonlogarithmic design, and the hybrid design.…”

Section: Introductionmentioning

confidence: 99%

“…The unsigned logarithmic ILM-AA multiplier with an approximate adder, proposed by Ansari et al [ 59 ], further improves the error characteristics of the previously proposed designs. The dynamic range approximate logarithmic multiplier DR-ALM5 proposed by Yin et al [ 57 ] dynamically truncates the input operands and thus uses smaller bit-width logarithmic and antilogarithmic converters and adders to generate the product. Pilipović et al [ 61 ] proposed an approximate logarithmic multiplier with two-stage operand trimming (TL16-8/4), which trims the least significant parts of the input operands in the first stage and the mantissas of the obtained operands’ approximations in the second stage.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

An Approximate GEMM Unit for Energy-Efficient Object Detection

Pilipović

Risojević

Božič

et al. 2021

Sensors

View full text Add to dashboard Cite

Edge computing brings artificial intelligence algorithms and graphics processing units closer to data sources, making autonomy and energy-efficient processing vital for their design. Approximate computing has emerged as a popular strategy for energy-efficient circuit design, where the challenge is to achieve the best tradeoff between design efficiency and accuracy. The essential operation in artificial intelligence algorithms is the general matrix multiplication (GEMM) operation comprised of matrix multiplication and accumulation. This paper presents an approximate general matrix multiplication (AGEMM) unit that employs approximate multipliers to perform matrix–matrix operations on four-by-four matrices given in sixteen-bit signed fixed-point format. The synthesis of the proposed AGEMM unit to the 45 nm Nangate Open Cell Library revealed that it consumed only up to 36% of the area and 25% of the energy required by the exact general matrix multiplication unit. The AGEMM unit is ideally suited to convolutional neural networks, which can adapt to the error induced in the computation. We evaluated the AGEMM units’ usability for honeybee detection with the YOLOv4-tiny convolutional neural network. The results implied that we can deploy the AGEMM units in convolutional neural networks without noticeable performance degradation. Moreover, the AGEMM unit’s employment can lead to more area- and energy-efficient convolutional neural network processing, which in turn could prolong sensors’ and edge nodes’ autonomy.

show abstract