FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3191560475> ?p ?o ?g. }

• W3191560475 abstract "Recent work demonstrated the promise of using resistive random access memory (ReRAM) as an emerging technology to perform inherently parallel analog domain in-situ matrix-vector multiplication—the intensive and key computation in deep neural networks (DNNs). One key problem is the weights that are signed values. However, in a ReRAM crossbar, weights are stored as conductance of the crossbar cells, and the in-situ computation assumes all cells on each crossbar column are of the same sign. The current architectures either use two ReRAM crossbars for positive and negative weights (PRIME), or add an offset to weights so that all values become positive (ISAAC). Neither solution is ideal: they either double the cost of crossbars, or incur extra offset circuity. To better address this problem, we propose FORMS, a fine-grained ReRAM-based DNN accelerator with algorithm/hardware co-design. Instead of trying to represent the positive/negative weights, our key design principle is to enforce exactly what is assumed in the in-situ computation— ensuring that all weights in the same column of a crossbar have the same sign. It naturally avoids the cost of an additional crossbar. Such polarized weights can be nicely generated using alternating direction method of multipliers (ADMM) regularized optimization during the DNN training, which can exactly enforce certain patterns in DNN weights. To achieve high accuracy, we divide the crossbar into logical sub-arrays and only enforce this property within the fine-grained sub-array columns. Crucially, the small sub-arrays provides a unique opportunity for input zero-skipping, which can significantly avoid unnecessary computations and reduce computation time. At the same time, it also makes the hardware much easier to implement and is less susceptible to non-idealities and noise than coarse-grained architectures. Putting all together, with the same optimized DNN models, FORMS achieves 1.50× and 1.93× throughput improvement in terms of $frac{{GOPs}}{{s times m{m^2}}}$ and $frac{{GOPs}}{W}$ compared to ISAAC, and 1.12× ~2.4 × speed up in terms of frame per second over optimized ISAAC with almost the same power/area cost. Interestingly, FORMS optimization framework can even speed up the original ISAAC from 10.7 × up to 377.9×, reflecting the importance of software/hardware co-design optimizations." @default.
• W3191560475 created "2021-08-16" @default.
• W3191560475 creator A5007300551 @default.
• W3191560475 creator A5013881064 @default.
• W3191560475 creator A5016070401 @default.
• W3191560475 creator A5025596795 @default.
• W3191560475 creator A5030060072 @default.
• W3191560475 creator A5032487759 @default.
• W3191560475 creator A5034466262 @default.
• W3191560475 creator A5047215143 @default.
• W3191560475 creator A5067008084 @default.
• W3191560475 creator A5071802262 @default.
• W3191560475 creator A5080407824 @default.
• W3191560475 date "2021-06-01" @default.
• W3191560475 modified "2023-10-06" @default.
• W3191560475 title "FORMS: Fine-grained Polarized ReRAM-based In-situ Computation for Mixed-signal DNN Accelerator" @default.
• W3191560475 cites W1578783943 @default.
• W3191560475 cites W2056507634 @default.
• W3191560475 cites W2140322178 @default.
• W3191560475 cites W2159543793 @default.
• W3191560475 cites W2162651880 @default.
• W3191560475 cites W2164819109 @default.
• W3191560475 cites W2233116163 @default.
• W3191560475 cites W2254450385 @default.
• W3191560475 cites W2295652899 @default.
• W3191560475 cites W2297924971 @default.
• W3191560475 cites W2508602506 @default.
• W3191560475 cites W2518281301 @default.
• W3191560475 cites W2536491920 @default.
• W3191560475 cites W2539856142 @default.
• W3191560475 cites W2592800647 @default.
• W3191560475 cites W2593769476 @default.
• W3191560475 cites W2606722458 @default.
• W3191560475 cites W2613989746 @default.
• W3191560475 cites W2614813641 @default.
• W3191560475 cites W2725159389 @default.
• W3191560475 cites W2736591611 @default.
• W3191560475 cites W2740220207 @default.
• W3191560475 cites W2742513757 @default.
• W3191560475 cites W2783393355 @default.
• W3191560475 cites W2791175987 @default.
• W3191560475 cites W2798554798 @default.
• W3191560475 cites W2807750997 @default.
• W3191560475 cites W2884159178 @default.
• W3191560475 cites W2902132251 @default.
• W3191560475 cites W2912012512 @default.
• W3191560475 cites W2913104037 @default.
• W3191560475 cites W2928560789 @default.
• W3191560475 cites W2946047477 @default.
• W3191560475 cites W2949674408 @default.
• W3191560475 cites W2949989598 @default.
• W3191560475 cites W2962818002 @default.
• W3191560475 cites W2962851801 @default.
• W3191560475 cites W2963027573 @default.
• W3191560475 cites W2963396654 @default.
• W3191560475 cites W2963944087 @default.
• W3191560475 cites W2964253307 @default.
• W3191560475 cites W2971533524 @default.
• W3191560475 cites W2979577066 @default.
• W3191560475 cites W2980104813 @default.
• W3191560475 cites W2989569745 @default.
• W3191560475 cites W3013407975 @default.
• W3191560475 cites W3014166398 @default.
• W3191560475 cites W3015980402 @default.
• W3191560475 cites W3021906994 @default.
• W3191560475 cites W3042493405 @default.
• W3191560475 cites W3080915835 @default.
• W3191560475 cites W3091885635 @default.
• W3191560475 cites W3104393472 @default.
• W3191560475 cites W3139203094 @default.
• W3191560475 cites W3162727751 @default.
• W3191560475 cites W4212788319 @default.
• W3191560475 cites W4234738201 @default.
• W3191560475 cites W4292363360 @default.
• W3191560475 doi "https://doi.org/10.1109/isca52012.2021.00029" @default.
• W3191560475 hasPublicationYear "2021" @default.
• W3191560475 type Work @default.
• W3191560475 sameAs 3191560475 @default.
• W3191560475 citedByCount "13" @default.
• W3191560475 countsByYear W31915604752022 @default.
• W3191560475 countsByYear W31915604752023 @default.
• W3191560475 crossrefType "proceedings-article" @default.
• W3191560475 hasAuthorship W3191560475A5007300551 @default.
• W3191560475 hasAuthorship W3191560475A5013881064 @default.
• W3191560475 hasAuthorship W3191560475A5016070401 @default.
• W3191560475 hasAuthorship W3191560475A5025596795 @default.
• W3191560475 hasAuthorship W3191560475A5030060072 @default.
• W3191560475 hasAuthorship W3191560475A5032487759 @default.
• W3191560475 hasAuthorship W3191560475A5034466262 @default.
• W3191560475 hasAuthorship W3191560475A5047215143 @default.
• W3191560475 hasAuthorship W3191560475A5067008084 @default.
• W3191560475 hasAuthorship W3191560475A5071802262 @default.
• W3191560475 hasAuthorship W3191560475A5080407824 @default.
• W3191560475 hasBestOaLocation W31915604752 @default.
• W3191560475 hasConcept C11413529 @default.
• W3191560475 hasConcept C121332964 @default.
• W3191560475 hasConcept C173608175 @default.
• W3191560475 hasConcept C17525397 @default.
• W3191560475 hasConcept C175291020 @default.
• W3191560475 hasConcept C182019814 @default.

• next