Mitigating the Effects of RRAM Process Variation on the Accuracy of Artificial Neural Networks

Fritscher, Markus; Knödtel, Johannes; Mallah, Maen; Pechmann, Stefan; Quesada, Emilio Pérez-Bosch; Rizzi, Tommaso; Wenger, Christian; Reichenbach, Marc

doi:10.1007/978-3-031-04580-6_27

Cited by 5 publications

(2 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore, unlike the quantization constraint, we should closely control the write variations in any future design for an acceptable basecaller. Fortunately, some previous works [22,37,100] propose mitigation techniques that, when combined, can provide us with reasonable (e.g., amount of ≤ 10%) write variation. From now on, we consider only up to 10% write variation (as defined in Section 2.3) in our evaluations.…”

Section: Accuracy For Write Variationsmentioning

confidence: 99%

Demeter: A Fast and Energy-Efficient Food Profiler Using Hyperdimensional Computing in Memory

et al. 2022

View full text Add to dashboard Cite

Food profiling is an essential step in any food monitoring system needed to prevent health risks and potential frauds in the food industry. Significant improvements in sequencing technologies are pushing food profiling to become the main computational bottleneck. State-of-the-art profilers are unfortunately too costly for food profiling. Our goal is to design a food profiler that solves the main limitations of existing profilers, namely (1) working on massive data structures and (2) incurring considerable data movement, for a real-time monitoring system. To this end, we propose Demeter, the first platform-independent framework for food profiling. Demeter overcomes the first limitation through the use of hyperdimensional computing (HDC) and efficiently performs the accurate few-species classification required in food profiling. We overcome the second limitation by the use of an in-memory hardware accelerator for Demeter (named Acc-Demeter) based on memristor devices. Acc-Demeter actualizes several domain-specific optimizations and exploits the inherent characteristics of memristors to improve the overall performance and energy consumption of Acc-Demeter. We compare Demeter's accuracy with other industrial food profilers using detailed software modeling. We synthesize Acc-Demeter's required hardware using UMC's 65nm library by considering an accurate PCM model based on silicon-based prototypes. Our evaluations demonstrate that Acc-Demeter achieves a (1) throughput improvement of 192× and 724× and (2) memory reduction of 36× and 33× compared to Kraken2 and MetaCache (2 state-of-the-art profilers), respectively, on typical food-related databases. Demeter maintains an acceptable profiling accuracy (within 2% of existing tools) and incurs a very low area overhead.

show abstract

Section: Accuracy For Write Variationsmentioning

confidence: 99%

Demeter: A Fast and Energy-Efficient Food Profiler Using Hyperdimensional Computing in Memory

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Mehmood et al [39], [40] presented an energy-efficient fault-tolerant scheme to upgrade the accuracy of WBAN. Fritscher et al [41] showed how a faultaware training act on the reaction of a network changeability. In the present article, we studied the properties of faulttolerant resolving sets and fault-tolerant metric dimensions of three r-level wheel related networks r-level anti-web wheel, gear and Helm network.…”

Section: Introductionmentioning

confidence: 99%

Studies of Multilevel Networks via Fault-Tolerant Metric Dimensions

2022

View full text Add to dashboard Cite

A subset T of the vertex set of a network G is called a resolving set for G if each pair of vertices of G have distinct representations with respect to T . A resolving set B among all the resolving sets of a network G is called a fault-tolerant resolving set if B \ {t} is as well a resolving set for each vertex t ∈ B . A fault-tolerant resolving set B of a network G which contains minimum number of vertices is called a fault-tolerant metric basis. The cardinality of a fault-tolerant metric basis is called faulttolerant metric dimension. This concept is widely used to find the integral solution of the problems existing in different disciplines of computer science and chemistry such as linear optimization problems, robot navigation, operation research problems, sensor networking, classification of chemical compounds, drug discoveries, source localization, embedding biological sequence data, detecting network motifs, comparing the interconnected networks and image processing. In this paper, we compute the fault-tolerant metric dimensions of three wheel related networks called by r-level anti-web wheel AW W (n,r) , r-level Helm H (n,r) and r-level anti-web gear AW J (2n,r) networks in the form of different algebraic expressions consisting of the integral numbers n and r. At the end we discussed a simple method for finding the faulttolerant metric dimensions and fault-tolerant resolving sets of a r-level wheel related network. We also discussed the importance of these networks in navigation.INDEX TERMS Fault-tolerant metric dimensions; Fault-tolerant resolving set; r-level anti web wheel network; r-level anti web gear network; r-level Helm network.

show abstract

Prototyping Reconfigurable RRAM-Based AI Accelerators Using the RISC-V Ecosystem and Digital Twins

Fritscher,

Veronesi,

Baroni

et al. 2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Mitigating the Effects of RRAM Process Variation on the Accuracy of Artificial Neural Networks

Cited by 5 publications

References 24 publications

Demeter: A Fast and Energy-Efficient Food Profiler Using Hyperdimensional Computing in Memory

Demeter: A Fast and Energy-Efficient Food Profiler Using Hyperdimensional Computing in Memory

Studies of Multilevel Networks via Fault-Tolerant Metric Dimensions

Prototyping Reconfigurable RRAM-Based AI Accelerators Using the RISC-V Ecosystem and Digital Twins

Contact Info

Product

Resources

About