Limits...
Comparative Study on Statistical-Variation Tolerance Between Complementary Crossbar and Twin Crossbar of Binary Nano-scale Memristors for Pattern Recognition.

Truong SN, Shin S, Byeon SD, Song J, Mo HS, Min KS - Nanoscale Res Lett (2015)

Bottom Line: In this comparative study, 10 greyscale images and 26 black-and-white alphabet characters are tested using the circuit simulator to compare the recognition rate with varying statistical variation and correlation parameters.As with the simulation results of 10 greyscale image recognitions, the twin crossbar shows better recognition rate by 4 % on average than the complementary one, when the inter-array correlation = 1 and intra-array correlation = 0.When the inter-array correlation = 1 and intra-array correlation = 1, the twin architecture is better by 6 % on average than the complementary one.By summary, we can conclude that the twin crossbar is more robust than the complementary one under the same amounts of statistical variation and correlation.

View Article: PubMed Central - PubMed

Affiliation: School of Electrical Engineering, Kookmin University, 77, Jeongneung-ro, Seongbuk-gu, Seoul, 136-702, South Korea. sontn@kookmin.ac.kr.

ABSTRACT
This paper performs a comparative study on the statistical-variation tolerance between two crossbar architectures which are the complementary and twin architectures. In this comparative study, 10 greyscale images and 26 black-and-white alphabet characters are tested using the circuit simulator to compare the recognition rate with varying statistical variation and correlation parameters.As with the simulation results of 10 greyscale image recognitions, the twin crossbar shows better recognition rate by 4 % on average than the complementary one, when the inter-array correlation = 1 and intra-array correlation = 0. When the inter-array correlation = 1 and intra-array correlation = 1, the twin architecture can recognize better by 5.6 % on average than the complementary one.Similarly, when the inter-array correlation = 1 and intra-array correlation = 0, the twin architecture can recognize 26 alphabet characters better by 4.5 % on average than the complementary one. When the inter-array correlation = 1 and intra-array correlation = 1, the twin architecture is better by 6 % on average than the complementary one. By summary, we can conclude that the twin crossbar is more robust than the complementary one under the same amounts of statistical variation and correlation.

No MeSH data available.


Memristor crossbar array write schemes. a 1/2VDD write scheme. b 1/3VDD write scheme [17]
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4608951&req=5

Fig6: Memristor crossbar array write schemes. a 1/2VDD write scheme. b 1/3VDD write scheme [17]

Mentions: The training of the crossbar circuit means changing memristor resistance value to HRS or LRS. Here, we can use the 1/2VDD write scheme or 1/3VDD write scheme [17] in training memristors to have the target resistance values of HRS and LRS in this paper. Figure 6a shows the 1/2VDD write scheme, where the selected cell is applied by VDD and GND. Here, the unselected cells on the same row or column with the selected cell are driven by 1/2VDD. If the resistance change due to this 1/2VDD is much smaller than the resistance change due to the full VDD, the unselected cells with 1/2VDD can keep their resistance values unchanged during the training process. If the unselected cells which should be driven by 1/2VDD are very susceptible to this small voltage of 1/2VDD, we can use the 1/3VDD write scheme, as shown in Fig. 6b. In this figure, the selected cell is applied by VDD and GND, like the selected cell in Fig. 6a. However, the unselected cells in Fig. 6b are driven by only 1/3VDD that is much smaller than the 1/2VDD in Fig. 6a [17]. By doing so, we can suppress the unwanted resistance change of the unselected cells in Fig. 6b better than those in Fig. 6a.Fig. 6


Comparative Study on Statistical-Variation Tolerance Between Complementary Crossbar and Twin Crossbar of Binary Nano-scale Memristors for Pattern Recognition.

Truong SN, Shin S, Byeon SD, Song J, Mo HS, Min KS - Nanoscale Res Lett (2015)

Memristor crossbar array write schemes. a 1/2VDD write scheme. b 1/3VDD write scheme [17]
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4608951&req=5

Fig6: Memristor crossbar array write schemes. a 1/2VDD write scheme. b 1/3VDD write scheme [17]
Mentions: The training of the crossbar circuit means changing memristor resistance value to HRS or LRS. Here, we can use the 1/2VDD write scheme or 1/3VDD write scheme [17] in training memristors to have the target resistance values of HRS and LRS in this paper. Figure 6a shows the 1/2VDD write scheme, where the selected cell is applied by VDD and GND. Here, the unselected cells on the same row or column with the selected cell are driven by 1/2VDD. If the resistance change due to this 1/2VDD is much smaller than the resistance change due to the full VDD, the unselected cells with 1/2VDD can keep their resistance values unchanged during the training process. If the unselected cells which should be driven by 1/2VDD are very susceptible to this small voltage of 1/2VDD, we can use the 1/3VDD write scheme, as shown in Fig. 6b. In this figure, the selected cell is applied by VDD and GND, like the selected cell in Fig. 6a. However, the unselected cells in Fig. 6b are driven by only 1/3VDD that is much smaller than the 1/2VDD in Fig. 6a [17]. By doing so, we can suppress the unwanted resistance change of the unselected cells in Fig. 6b better than those in Fig. 6a.Fig. 6

Bottom Line: In this comparative study, 10 greyscale images and 26 black-and-white alphabet characters are tested using the circuit simulator to compare the recognition rate with varying statistical variation and correlation parameters.As with the simulation results of 10 greyscale image recognitions, the twin crossbar shows better recognition rate by 4 % on average than the complementary one, when the inter-array correlation = 1 and intra-array correlation = 0.When the inter-array correlation = 1 and intra-array correlation = 1, the twin architecture is better by 6 % on average than the complementary one.By summary, we can conclude that the twin crossbar is more robust than the complementary one under the same amounts of statistical variation and correlation.

View Article: PubMed Central - PubMed

Affiliation: School of Electrical Engineering, Kookmin University, 77, Jeongneung-ro, Seongbuk-gu, Seoul, 136-702, South Korea. sontn@kookmin.ac.kr.

ABSTRACT
This paper performs a comparative study on the statistical-variation tolerance between two crossbar architectures which are the complementary and twin architectures. In this comparative study, 10 greyscale images and 26 black-and-white alphabet characters are tested using the circuit simulator to compare the recognition rate with varying statistical variation and correlation parameters.As with the simulation results of 10 greyscale image recognitions, the twin crossbar shows better recognition rate by 4 % on average than the complementary one, when the inter-array correlation = 1 and intra-array correlation = 0. When the inter-array correlation = 1 and intra-array correlation = 1, the twin architecture can recognize better by 5.6 % on average than the complementary one.Similarly, when the inter-array correlation = 1 and intra-array correlation = 0, the twin architecture can recognize 26 alphabet characters better by 4.5 % on average than the complementary one. When the inter-array correlation = 1 and intra-array correlation = 1, the twin architecture is better by 6 % on average than the complementary one. By summary, we can conclude that the twin crossbar is more robust than the complementary one under the same amounts of statistical variation and correlation.

No MeSH data available.