Ferroelectric Memristive Networks for Dimensionality Reduction: A Process for Effectively Classifying Cancer Datasets

Abstract

In this work, a copper-doped (5%) zinc oxide (Cu:ZnO) ferroelectric materials-based memristor model was realized and it was employed to develop principal component analysis (PCA), a data dimension reduction technique. The developed PCA was utilized to efficaciously classify breast cancer datasets, which are considered as complex and big volumes of data. It was found that the controllable memristance variations were analogous to the weight modulations in the implemented neural network-based learning systems. Sanger’s rule was utilized to achieve unsupervised online learning in order to generate the principal components. On one side, the developed memristor-based PCA network was found to be effective to isolate distinct breast cancer classes with a high classification accuracy of 97.77% and the error in the classification of malignant cases as benign of 0.529%, a significantly low value. On the other side, the power dissipation was found to be 0.27 µW, which suggests the proposed memristive network is suitable for low-power applications. Further, a comparison was established with other existing non-memristor and non-PCA-based data classification systems. Furthermore, the devised less complex equations to implement PCA on this memristive crossbar array could be employed to implement any neural network algorithm.

Memristor-based signal processing for edge computing

“…High classification accuracy of 97.1% indicates the outstanding performance of such a memristor-based PCA. With ferroelectric materialbased memristors, Raj et al [60] further suggested that an improved performance of PCA can be achieved in simulations. Moreover, techniques, such as online learning or hybrid training [61] , could accommodate device nonidealities, and thus they can be developed and employed for memristor-based signal processing systems to further improve performance.…”

“…N/A Signal transform DFT [40][41][42] E Time-frequency transformation [40] and speech recognition [42] N/A 10 in speed, 109.8 in power efficiency [40] DCT [5,45] E Image compression and processing [5] Energy efficiency: 119.7 TOPs 1 W 1 [5] N/A DWT [44] S Image compression [44] Energy: 6.4 nJ/image Time: 15 s/image [44] 11 in energy efficiency, 1.28 in speed [44] Signal encoding CS [46,48,49,51] E Image compression and reconstruction [51] Power dissipation: 16.2 mW/read [51] 50 in power consumption [51] SC [47,[54][55][56] E Sparse representation of natural images [47] Energy: 719 J/image Time: 0.036 s/image [47] 16 in energy consumption [47] Component analysis PCA [58][59][60] E Classification of breast cancer [60] Power dissipation: 0.27 W/feature [60] N/A ICA [62][63][64] E Blind image source separation [64] N/A N/A Classification and regression N/A SVM [66,67] S Wake-up system [66] Energy: 0.7 nJ for potentiation, 0.5 pJ for depression [66] N/A SLP [68,…”