Accuracy evaluation of a trained neural network by energy efficient approximate 4:2 compressor. (June 2021)

• Record Type: Journal Article
• Authors: Maddisetti, Lavanya; Senapati, Ranjan K.; JVR, Ravindra
• Is Part Of: Computers & electrical engineering. Volume 92(2021)
• Publication Date: 2021-06
• Subjects: 00-01 -- 99-00; Approximate circuits -- CMOS -- Compressor -- Multiplier -- Neural network; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2021.107137
• Languages: English
• ISSNs: 0045-7906

Abstract

Abstract: Automation techniques and machine learning algorithms are playing a crucial role in almost all fields in recent times. In this research, a 4:2 compressor circuit is approximated using the probabilistic pruning technique. An artificial neural network is designed for the proposed 4:2 compressor and is trained to obtain the train and test accuracies. The neural network with equal train and test accuracies has been considered as the best approximate circuit. The training of the neural network has been performed using a supervised machine learning algorithm by applying truth table of the proposed approximate 4:2 compressor as the dataset. The proposed compressor has only 19 transistors and consumes less energy i.e., 0.2015 nJ with less silicon area of 14.36 um 2 . The performance of the Dadda multiplier is improved by replacing the proposed approximate 4:2 compressor into its partial product reduction stage. Graphical abstract: Highlights: Probabilistic pruning type of approximation is applied on the exact 4:2 compressor An ANN designed with binary weights and inputs from S u m A of proposed 4:2 compressor. An environment established for agent to gain rewards for selecting best compressor