The proliferation of information has facilitated international accessibility, which calls for escalating quantities of energy for knowledge storage and processing functions. In recent times, there was an increase in analysis within the discipline of neuromorphic electronics, which attracts inspiration from organic neurons and synapses. These electronics possess the flexibility to carry out in-memory computing, which helps alleviate the restrictions imposed by the ‘von Neumann bottleneck’ that exists between the reminiscence and processor within the conventional von Neumann structure. By leveraging their multi-bit non-volatility, traits that mimic biology, and Kirchhoff’s legislation, neuromorphic electronics provide a promising resolution to cut back the facility consumption in processing vector–matrix multiplication duties. Amongst all the present nonvolatile reminiscence applied sciences, NAND flash reminiscence is among the best built-in options for the storage of huge volumes of information. This work offers a complete overview of the latest developments in neuromorphic computing based mostly on NAND flash reminiscence. Neuromorphic architectures utilizing NAND flash reminiscence for off-chip studying are offered with varied quantization ranges of enter and weight. Subsequent, neuromorphic architectures for on-chip studying are offered utilizing customary backpropagation and suggestions alignment algorithms. The array structure, operation scheme, and electrical traits of NAND flash reminiscence are mentioned with a give attention to using NAND flash reminiscence in varied neural community constructions. Moreover, the discrepancy of array structure between on-chip studying and off-chip studying is addressed. This evaluate article offers a basis for understanding the neuromorphic computing based mostly on the NAND flash reminiscence and strategies to put it to use based mostly on software necessities.
