ReRAM CiM Fluctuation Pattern Classification by CNN Trained on Artificially Created Dataset

3D Parallel ReRAM Computation-in-Memory for Hyperdimensional Computing

“…For large-scale algorithms such as Machine Learning, the cost of data transfer is a major barrier, known as the von Neumann bottleneck, and CiM is seen as a promising way to solve this problem. In addition, using ReRAM, one of the emerging Non-Volatile Memories (NVMs), grants some advantages such as energy efficiency [9][10][11][12][13].…”

Comprehensive Analysis of Read Fluctuations in ReRAM CiM by Using Fluctuation Pattern Classifier

“…In recent years, conventional Neumann-type computer architecture faces issues of energy consumption and large latency by transporting data from memory to the processor. To tackle this problem, Computation-in-Memory (CiM) has been proposed [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16]. It computes multiply-accumulate (MAC) operations in memory array without data transportation.…”

“…It computes multiply-accumulate (MAC) operations in memory array without data transportation. As a memory of CiM, several types of memories have been considered, i.e., static random access memory (SRAM) [1], resistive RAM (ReRAM) [2][3][4][5][6][7][8][9][10][11][12], phase change memory (PCM) [13], and ferroelectric field effect transistor (FeFET) [14][15]. Although volatile memory-based CiM has advantages in some points such as low error rate and process cost, non-volatile memory-based CiM (nvCiM) is a good option for edge application because of its analog characteristics and energy-efficiency.…”

“…Thus, memory non-idealities, including write variation and read-disturb [17] are the large issues in ReRAM CiM [16]. In addition, ReRAM has been suffered from its conductance fluctuation [2][3][4][5][6][7][18][19][20][21][22][23][24][25][26]. The main fluctuation pattern has been considered the random telegraph noise (RTN) conventionally, but there are still other fluctuation patterns are reported [2,[18][19][20][21][22]26].…”

“…In addition, ReRAM has been suffered from its conductance fluctuation [2][3][4][5][6][7][18][19][20][21][22][23][24][25][26]. The main fluctuation pattern has been considered the random telegraph noise (RTN) conventionally, but there are still other fluctuation patterns are reported [2,[18][19][20][21][22]26]. G. González-Cordero et al (2021) adopted self-organization map (SOM) to fluctuating signal analysis [26], but it does not work properly on our measured signals.…”