Search

Resistive in-memory computing (IMC) architectures currently lag behind SRAM IMCs and digital accelerators in both energy efficiency and compute density due to their low compute accuracy. This article proposes the use of signal-to-noise-plus-distortion ratio (SNDR) to quantify the compute accuracy of IMCs and identify the device, circuit, and architectural parameters that affect it. We further analyze the fundamental limits on the SNDR of magnetoresistive random access memory (MRAM-), resistive random access memory (ReRAM-), and ferroelectric field effect transistor (FeFET)-based IMCs employing parameter variation and noise models that were validated against measured results from a recent MRAM-based IMC prototype in a 22 nm process. At high-output signal magnitude, we can find that the maximum achievable SNDR is limited by the pre-analog-to-digital-converter (ADC) array nonidealities, such as the conductance variations (CVs), parasitic resistances, and current mirror mismatch (MM), whereas the ADC thermal (AT) noise limits the SNDR at small signal magnitudes. Furthermore, for large dot-product (DP) dimensions ( $N > 50$ ), the maximum achievable SNDR is highest for FeFET, followed by ReRAM and then MRAM. Finally, the increase in conductance contrast ( ${g_ {\text {ON}} }/ {g_ {\text {OFF}} }$ ) enhances the maximum achievable SNDR only until it reaches a value of approximately 12. ReRAMs and FeFETs demonstrate high energy efficiencies while achieving high SNDR, as their low conductance values lead to lower currents and lower noise due to wire parasitics. In all cases, across all three device types, DP dimension, ADC precision, and conductance contrast, the maximum achievable SNDR is found to be in the range of 18–22 dB, barely meeting the minimum needed for achieving an inference accuracy close to an equivalent fixed-point digital architecture. Finally, we demonstrate a network-level accuracy of 84.5% when mapping an ResNet-20 (CIFAR-10) by ReRAM-based architecture at a SNDR of 22 dB, in which MRAM- and FeFET-based architectures cannot realize. This result clearly implies the need for other approaches, e.g., algorithmic- and learning-based methods, to improve the inference accuracy of resistive IMC architectures.

In-memory computing (IMC) architectures have emerged as a compelling platform to implement energy-efficient machine learning (ML) systems. However, today, the energy efficiency gains provided by IMC designs seem to be leveling off and it is not clear what the limiting factors are. The conceptual complexity of IMCs combined with the absence of a rigorous benchmarking methodology makes it difficult to gauge progress and identify bottlenecks in this exciting field. This article presents a benchmarking methodology for IMCs comprising: 1) a compositional view of IMCs that enables one to parse an IMC design into its canonical components; 2) a set of benchmarking metrics to quantify the performance, efficiency, and accuracy of IMCs; and 3) a strategy for analyzing the reported IMC data and metrics. We apply the proposed benchmarking methodology on an extensive database of IMC metrics extracted from > 70 IC designs published since 2018, in order to infer and comprehend trends in this area. Our benchmarking effort indicates: 1) SRAM-based IMCs show a clear win in terms of energy efficiency and compute density over digital accelerators at the bank level but the energy efficiency gap reduces dramatically when comparing at the processor level; 2) eNVM-based IMCs lag behind SRAM-based IMCs in terms of both energy efficiency and compute density, and surprisingly lag digital accelerators in terms of compute density; 3) the compute (bank-level) accuracy of IMCs, though a critical metric, is pervasively neglected in publications as is the energy versus accuracy tradeoff inherent to IMCs.

This study attempts to develop a model that forecasts precise data on traffic flow. Everything that can impact the flow of traffic on the road is referred to as the traffic environment, including traffic signals, accidents, rallies, and even road repairs that could result in a traffic bottleneck. The driver or passenger can make an informed choice if they have prior knowledge about the vehicle crowd close to the area that will have the greatest impact on traffic. Additionally, it can be utilised in driverless vehicles, which are the automobiles of the future. Today’s traffic is increasing tremendously, and big data transportation concepts are becoming more popular. We are motivated to develop a machine learning model that forecasts traffic flow because the present prediction techniques and models are still insufficient for use in practical applications. The amount of data available to forecast traffic flow is so enormous that it is awkward and laborious. In this work, we intended to evaluate the data for the transportation with significantly less complexity using machine learning and deep learning methods. The user will be informed of the projected information and the constructed machine learning model will predict the traffic flow.

Background: Both hypertension (HTN) and tobacco addiction pose a threat to the health, environment, and socioeconomic status (SES) of the people. When tobacco use disorder exists in people with HTN, it hastens the disease progress and causes early complications. The present study aimed to study the knowledge and practice of tobacco addiction in patients with HTN and find out the correlates of knowledge and practice of tobacco addiction. Methods: A cross-sectional study was conducted for a period of one year in the Jamnagar District of Western Gujarat, India. Out of total 400 samples, 50% were collected from the five selected Community Health Centers (CHCs) by random sampling and the rest from non-communicable disease (NCD) clinics at the tertiary care hospital of the district. Findings: Most of the patients were in their fifties or above (67.0%), women (57.0%), and married (86.5%). Only 12.0% had awareness about the hazards of tobacco addiction and the prevalence of tobacco addiction was 11%. It was found that use of smokeless tobacco (SLT) among samples was the highest (72.7%) followed by dual consumption, i.e., SLT and smoking (20.5%). The frequency of consumption was ≥ 5 times/day in 54.5%, 70.4% were addicted for > 10 years, and only 15.9% had ever tried to quit tobacco while only 11.4% had successfully quit it. It was seen that those who were aged < 50 years, men, literates, employed, and those belonging to higher SES and urban residence had better knowledge of the health hazards. This was found to have significant statistical association. Conclusion: The results provide valuable insight into the tobacco addiction in patients with HTN on which non-pharmacological treatment of HTN can be based.

The energy and delay reductions from CMOS scaling have stagnated, motivating the search for a CMOS replacement. Spintronic devices are one of the promising beyond-CMOS alternatives. However, they exhibit high switching error rates of 1% or more when operated at energy and delay comparable to CMOS, rendering them incompatible with the deterministic nature of digital implementations. In this paper, we employ a Shannon-inspired model of computation to enhance the tolerance of all-spin logic (ASL)-based implementations to gate-level switching errors. We develop the logic-level path delay reallocation techniques to shape the output error statistics and propose a novel error compensation scheme to achieve 1000× higher tolerance to device-level switching errors while maintaining the classification accuracy of an ASL-based support vector machine (SVM) classifier.

Computational scaling beyond silicon electronics based on Moore's law requires the adoption of alternate state variables such as electronic spin. Multiple research efforts are underway exploring both Boolean and non-Boolean design space using spin devices in order to make their energy and delay benefits competitive to CMOS. In this paper, we propose spin channel networks (SCN), where the exponential decay property of spin current along the spin channel is exploited to achieve energy-efficient dot product implementation for inference applications. As the use of exponentially decaying spin current for analog computation enforces severe locality constraints, we employ adaptive boosting to design an ensemble of tiny SCNs that work in unison to solve any binary classification task. Such boosted SCNs achieve up to 112× and 14× higher energy efficiency over conventional all-spin-logic-based and 20-nm CMOS designs, respectively.

Numerical wind mapping is currently the wind power industry’s standard for preliminary assessments for sites of good wind resources. Of the various available numerical models, numerical weather prediction (NWP) models are best suited for modeling mesoscale wind flows across small islands. In this study, the Weather Research and Forecast (WRF) NWP model was optimized for simulating the wind resources of the Caribbean islands of Trinidad and Tobago in terms of spin-up period for developing mesoscale features, the input initial and boundary conditions, and the planetary boundary layer (PBL) parameterizations. Hourly model simulations of wind speed and wind direction for a one-month period were compared with corresponding 10 m level wind observations. The National Center for Environmental Prediction (NCEP)-Department of Energy (DOE) reanalysis of 1.875° horizontal resolution was found to be better suited to provide initial and boundary conditions (ICBCs) over the 1° resolution NCEP final analysis (FNL); 86% of modeled wind speeds were within ±2 m/s of measured values at two locations when the coarse resolution NCEP reanalysis was used as compared with 55–64% of modeled wind speeds derived from FNL. Among seven PBL schemes tested, the Yonsei University PBL scheme with topographic drag enabled minimizes the spatial error in wind speed (mean bias error +0.16 m/s, root-mean-square error 1.53 m/s and mean absolute error 1.21 m/s) and is capable of modeling the bimodal wind speed histogram. These sensitivity tests provide a suitable configuration for the WRF model for mapping the wind resources over Trinidad and Tobago, which is a factor in developing a wind energy sector in these islands.