Memory Mode Research Articles

Logic-Compatible Embedded DRAM Architecture for Multifunctional Digital Storage and Compute-in-Memory

The compute-in-memory (CIM) which embeds computation inside memory is an attractive scheme to circumvent von Neumann bottlenecks. This study proposes a logic-compatible embedded DRAM architecture that supports data storage as well as versatile digital computations. The proposed configurable memory unit operates in three modes: (1) memory mode in which it works as a normal dynamic memory, (2) logic–arithmetic mode where it performs bit-wise Boolean logic and full adder operations on two words stored within the memory array, and (3) convolution mode in which it executes digitally XNOR-and-accumulate (XAC) operation for binarized neural networks. A 1.0-V 4096-word × 8-bit computational DRAM implemented in a 45-nanometer CMOS technology performs memory, logic and arithmetic operations at 241, 229, and 224 MHz while consuming the energy of 7.92, 8.09, and 8.19 pJ/cycle. Compared with conventional digital computing, it saves energy and latency of the arithmetic operation by at least 47% and 46%, respectively. For VDD = 1.0 V, the proposed CIM unit performs two 128-input XAC operations at 292 MHz with an energy consumption of 20.8 pJ/cycle, achieving 24.6 TOPS/W. This marks at least 11.9× better energy efficiency and 38.8× better delay, thereby achieving at least 461× better energy-delay product than traditional 8-bit wide computing hardware.

Unique Coexistence of Two Resistive Switching Modes in a Memristor Device Enables Multifunctional Neuromorphic Computing Properties.

We report on hybrid memristor devices consisting of germanium dioxide nanoparticles (GeO2 NP) embedded within a poly(methyl methacrylate) (PMMA) thin film. Besides exhibiting forming-free resistive switching and an uncommon "ON" state in pristine conditions, the hybrid (nanocomposite) devices demonstrate a unique form of mixed-mode switching. The observed stopping voltage-dependent switching enables state-of-the-art bifunctional synaptic behavior with short-term (volatile/temporal) and long-term (nonvolatile/nontemporal) modes that are switchable depending on the stopping voltage applied. The short-term memory mode device is demonstrated to further emulate important synaptic functions such as short-term potentiation (STP), short-term depression (STD), paired-pulse facilitation (PPF), post-tetanic potentiation (PTP), spike-voltage-dependent plasticity (SVDP), spike-duration-dependent plasticity (SDDP), and, more importantly, the "learning-forgetting-rehearsal" behavior. The long-term memory mode gives additional long-term potentiation (LTP) and long-term depression (LTD) characteristics for long-term plasticity applications. The work shows a unique coexistence of the two resistive switching modes, providing greater flexibility in device design for future adaptive and reconfigurable neuromorphic computing systems at the hardware level.

Soil moisture prediction based on integrating the CEEMDAN decomposition technique with LSTM in the proximity of Prosopis Juliflora

Human-made adaptations in plant diversity, particularly the introduction of alien invasive plant species, can result in problems such as depletion of groundwater, extraction of natural water resources, and modifications in the local biodiversity. In addition to regulating ecosystems and accurate farming, precise forecasting of soil moisture in the vicinity of these exotic plant species can greatly assist in various other agricultural applications. In this study, we present an improved approach for predicting soil moisture levels at different depths in the vicinity of Prosopis Juliflora (IAPS) by integrating long short-term memory (LSTM) and complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN). CEEMDAN is employed as a preprocessing and denoising technique to decompose the initial soil moisture data into several intrinsic mode functions (IMF). After that, LSTM is used to predict all the IMF (Intrinsic Mode Function) parts from the CEEMDAN process. The individual prediction outcomes of each IMF are combined to obtain the final prediction results. Soil moisture measurements were collected continuously in the Coimbatore area of Tamil Nadu at multi-layer soil depths of 10 cm, 30 cm, 60 cm, and 100 cm. The provided data was utilized for training models and achieving optimal accuracy in forecasting soil moisture levels one month prior. The proposed CEEMDAN-LSTM model achieved the highest R2, ranging from 0.985 to 0.995 at various depths when compared to RNN, LSTM, EMD-LSTM, EEMD-LSTM models, and existing literature models. The results show that the proposed method works well at predicting soil moisture at different depths, which can help improve the planting pattern and schedule of native species to increase the area’s green cover.

Prediction of packing performance of pneumatic control valve based on stem mechanism model and VMD-SSA-LSTM

Abstract Packing performance is one of the most important parts affecting the control operation of pneumatic control valves. In order to avoid the failure caused by packing wear and aging of pneumatic control valves after long-term use, this paper constructs a valve stem mechanism model and combines the advantage of deep learning to predict the packing performance of pneumatic control valves based on Long Short-term Memory (LSTM) and Variable Mode Decomposition (VMD) and Sparrow Search Algorithm (SSA). Firstly, the historical data of the actuator is calculated based on the mechanism model of the valve stem. The time-friction force obtained by the calculation is the parameter set. After the parameter set is subject to the variational mode decomposition, the parameter set is imported into the SSA-LSTM prediction model for prediction, and the prediction data is obtained.

2D Reconfigurable Memory Device Enabled by Defect Engineering for Multifunctional Neuromorphic Computing.

In this era of artificial intelligence and Internet of Things, emerging new computing paradigms such as in-sensor and in-memory computing call for both structurally simple and multifunctional memory devices. Although emerging two-dimensional (2D) memory devices provide promising solutions, the most reported devices either suffer from single functionalities or structural complexity. Here, this work reports a reconfigurable memory device (RMD) based on MoS2/CuInP2S6 heterostructure, which integrates the defect engineering-enabled interlayer defects and the ferroelectric polarization in CuInP2S6, to realize a simplified structure device for all-in-one sensing, memory and computing. The plasma treatment-induced defect engineering of the CuInP2S6 nanosheet effectively increases the interlayer defect density, which significantly enhances the charge-trapping ability in synergy with ferroelectric properties. The reported device not only can serve as a non-volatile electronic memory device, but also can be reconfigured into optoelectronic memory mode or synaptic mode after controlling the ferroelectric polarization states in CuInP2S6. When operated in optoelectronic memory mode, the all-in-one RMD could diagnose ophthalmic disease by segmenting vasculature within biological retinas. On the other hand, operating as an optoelectronic synapse, this work showcases in-sensor reservoir computing for gesture recognition with high energy efficiency.

Synthesis of robust memory modes for linear quantum systems with unknown inputs

In this paper, the synthesis of robust memory modes for linear quantum passive systems in the presence of unknown inputs has been studied, aimed at facilitating secure storage and communication of quantum information. In particular, we can switch on decoherence-free (DF) modes in the storage stage by placing the poles on the imaginary axis via a coherent feedback control scheme, and these memory modes can further be simultaneously made robust against perturbations to the system parameters by minimizing the condition number associated with imaginary poles. The DF modes can also be switched off by tuning the controller parameters to place the poles in the left half of the complex plane in the writing/reading stage. We develop explicit algebraic conditions guiding the design of such a coherent quantum controller, which involves employing an augmented system model to counter the influence of unknown inputs. Examples are provided to illustrate the procedure of synthesizing robust memory modes for linear optical quantum systems.

A deep learning based communication traffic prediction approach for smart monitoring of distributed energy resources in virtual power plants

AbstractVirtual power plants (VPPs) have been widely recognized as a key enabler for energy system neutrality. The communication traffic of a VPP fundamentally indicates its activeness in interacting with the power system, thus providing a new dimension in depicting the behaviour characteristics of distributed energy resources in VPPs. Therefore, the prediction of communication traffic is significant in improving the control efficiency of VPPs. However, due to the involvement of numerous interactive agents characterised by both individual randomness and coordinative characteristics, traditional prediction models are no longer capable of fitting VPP communication traffic effectively. Therefore, a novel prediction model is introduced that enhances the prediction accuracy by integrating long short‐term memory (LSTM) and variational mode decomposition (VMD). This model employs VMD as the initial step for extracting the intrinsic modes from the traffic sequence, thereby mitigating the impact of incidental noise. Then, LSTM is applied to fit each intrinsic mode individually. Additionally, considering the outer influencing factors, the attention mechanism is incorporated. Finally, all sub‐prediction algorithms are neatly integrated as a whole prediction model. The proposed model is evaluated through simulation prediction using realistic VPP communication traffic data, and the results demonstrate its effectiveness.

Tracer dynamics in polymer networks: Generalized Langevin description.

Tracer diffusion in polymer networks and hydrogels is relevant in biology and technology, while it also constitutes an interesting model process for the dynamics of molecules in fluctuating, heterogeneous soft matter. Here, we systematically study the time-dependent dynamics and (non-Markovian) memory effects of tracers in polymer networks based on (Markovian) implicit-solvent Langevin simulations. In particular, we consider spherical tracer solutes at high dilution in regular, tetrafunctional bead-spring polymer networks and control the tracer-network Lennard-Jones (LJ) interactions and the polymer density. Based on the analysis of the memory (friction) kernels, we recover the expected long-time transport coefficients and demonstrate how the short-time tracer dynamics, polymer fluctuations, and the viscoelastic response are interlinked. Furthermore, we fit the characteristic memory modes of the tracers with damped harmonic oscillations and identify LJ contributions, bond vibrations, and slow network relaxations. Tuned by the LJ interaction parameter, these modes enter the kernel with an approximately linear to quadratic scaling, which we incorporate into a reduced functional form for convenient tracer memory interpolation and extrapolation. This eventually leads to highly efficient simulations utilizing the generalized Langevin equation, in which the polymer network acts as an additional thermal bath with a tunable intensity.

Photonic Synaptic Transistor with Memory Mode Switching for Neuromorphic Visual System

AbstractThe human retina is able to extract key feature information from a large amount of redundant visual information, which is the basis for efficient information processing in the human visual system. However, current retina‐inspired photonic synaptic devices lack fast noise filtering capabilities, limiting the speed of image preprocessing in neuromorphic visual systems. Here, a photonic synaptic transistor (PST) based on graphene/organic heterojunction that exhibits high photosensitivity and optically tunable synaptic characteristics from visible to near‐infrared (488–1310 nm) is demonstrated. The PST enables light‐intensity‐controlled memory‐free and long‐memory mode switching, allowing to achieve fast image noise filtering in a PST‐based vision sensor (processing times as low as 30 ms). In addition, image recognition in an artificial neural network connected by the PST, and the efficiency and accuracy of image recognition can be significantly improved by performing image noise filtering at the front‐end is demonstrated. This work provides the potential to improve the information processing speed of bio‐inspired neuromorphic visual systems and contribute to the development of machine vision applications.

The Effect of Memory and Presentation Mode in Melodic Error Detection

Previous research has variably indicated the role of working memory in error detection by which working memory played a role in rhythmic error detection but not melodic error detection. Here, we devised a longer melodic error detection task for college musicians in an auditory, rather than visual, condition using classical excerpts, which we compared to briefer visual and auditory control conditions. These tests were compared to performance on a test of verbal working memory (forward digit span test) and an experimenter-created tonal working memory test. The tonal working memory test was positively related to the forward digit span test, the melodic error detection, and the visual control but not to the auditory control. Performance on the error detection test was not significantly related to year in school, level of aural skills class, years of private piano, or level of group piano class. Our participants performed similarly on the aurally presented melodic error detection of classical excerpts and the briefer visual control but not on the briefer aural control. Among other variables, years of experience on a second instrument was a significant predictor of error detection skill. High familiarity ratings with a classical excerpt did not yield a relationship to error detection performance.

Full hardware implementation of neuromorphic visual system based on multimodal optoelectronic resistive memory arrays for versatile image processing

In-sensor and near-sensor computing are becoming the next-generation computing paradigm for high-density and low-power sensory processing. To fulfil a high-density and efficient neuromorphic visual system with fully hierarchical emulation of the retina and visual cortex, emerging multimodal neuromorphic devices for multi-stage processing and a fully hardware-implemented system with versatile image processing functions are still lacking and highly desirable. Here we demonstrate an emerging multimodal-multifunctional resistive random-access memory (RRAM) device array based on modified silk fibroin protein (MSFP), exhibiting both optoelectronic RRAM (ORRAM) mode featured by unique negative and positive photoconductance memory and electrical RRAM (ERRAM) mode featured by analogue resistive switching. A full hardware implementation of the artificial visual system with versatile image processing functions is realised for the first time, including ORRAM mode array for the in-sensor image pre-processing (contrast enhancement, background denoising, feature extraction) and ERRAM mode array for near-sensor high-level image recognition, which hugely improves the integration density, and simply the circuit design and the fabrication and integration complexity.

Sleep spindle architecture associated with distinct clinical phenotypes in older adults at risk for dementia

Sleep spindles are a hallmark of non-REM sleep and play a fundamental role in memory consolidation. Alterations in these spindles are emerging as sensitive biomarkers for neurodegenerative diseases of ageing. Understanding the clinical presentations associated with spindle alterations may help to elucidate the functional role of these distinct electroencephalographic oscillations and the pathophysiology of sleep and neurodegenerative disorders. Here, we use a data-driven approach to examine the sleep, memory and default mode network connectivity phenotypes associated with sleep spindle architecture in older adults (mean age = 66 years). Participants were recruited from a specialist clinic for early diagnosis and intervention for cognitive decline, with a proportion showing mild cognitive deficits on neuropsychological testing. In a sample of 88 people who underwent memory assessment, overnight polysomnography and resting-state fMRI, a k-means cluster analysis was applied to spindle measures of interest: fast spindle density, spindle duration and spindle amplitude. This resulted in three clusters, characterised by preserved spindle architecture with higher fast spindle density and longer spindle duration (Cluster 1), and alterations in spindle architecture (Clusters 2 and 3). These clusters were further characterised by reduced memory (Clusters 2 and 3) and nocturnal hypoxemia, associated with sleep apnea (Cluster 3). Resting-state fMRI analysis confirmed that default mode connectivity was related to spindle architecture, although directionality of this relationship differed across the cluster groups. Together, these results confirm a diversity in spindle architecture in older adults, associated with clinically meaningful phenotypes, including memory function and sleep apnea. They suggest that resting-state default mode connectivity during the awake state can be associated with sleep spindle architecture; however, this is highly dependent on clinical phenotype. Establishing relationships between clinical and neuroimaging features and sleep spindle alterations will advance our understanding of the bidirectional relationships between sleep changes and neurodegenerative diseases of ageing.

Within and between‐network functional connectivity disruptions in posterior cortical atrophy and dementia with Lewy bodies

AbstractBackgroundPosterior cortical atrophy (PCA) and dementia with Lewy bodies (DLB) show overlapping patterns of hypometabolism affecting posterior regions of the brain. Both syndromes show disruptions in functional connectivity, but it is unknown how patterns of disruption within and between networks compare between PCA and DLB.MethodThirty amyloid‐positive PCA, 27 amyloid‐positive DLB (DLB+), 24 amyloid‐negative DLB (DLB‐) patients and 30 cognitively normal amyloid‐negative (CN) individuals were recruited at Mayo Clinic, Rochester, MN, and underwent structural and resting state functional MRI. Spatially preprocessed data was analyzed using the CONN functional connectivity toolbox to explore nine networks: visual, sensorimotor, dorsal‐attention, cerebellar, frontoparietal, memory, language, salience, and default mode networks (DMN). Multivariate linear regression models corrected for multiple comparisons and adjusted for age and sex were fit to analyze within‐network connectivity (Fisher’s R‐to‐Z transformations). Between‐network connectivity was assessed with group level difference maps (two‐sample t‐test) corrected for false discovery rate at p<0.05.ResultMultivariate linear regression models showed reduced within‐network connectivity across all DMN components, sensorimotor, and dorsal‐attention networks, with trends for the visual network in PCA compared to CN. Both DLB‐ and DLB+ showed reduced within‐network connectivity for the cerebellar network, with DLB+ showing reduced connectivity in the posterior‐DMN, memory, and salience networks compared to CN. DLB‐ showed increased connectivity in dorsal‐attention and DMN, while DLB+ showing reduced connectivity in the memory network compared to PCA. Between‐network analysis showed an increase in cerebellar‐to‐sensorimotor connectivity in all three groups, with increase in cerebellar‐to‐dorsal‐attention connectivity in DLB‐ and PCA compared to CN. DLB‐ showed a decrease in posterior‐DMN‐to‐sensorimotor connectivity, while PCA showed reduced connectivity between all DMN components and dorsal‐attention‐to‐sensorimotor connectivity compared to CN. There was an increase in connectivity between all DMN components and a reduction in dorsal‐DMN‐to‐sensorimotor connectivity in both DLB‐ and DLB+ compared to PCA.ConclusionPCA was associated with connectivity disruptions across all DMN components, sensorimotor, dorsal‐attention and visual networks, while both DLB groups showed disruptions in cerebellar networks, with DLB+ showing further disruptions in posterior‐DMN, memory and salience networks. These findings suggest that both PCA and DLB have unique brain network disruptions, with amyloid status influencing connectivity in DLB.

Prediction accuracy improvement of pressure pulsation signals of reversible pump‐turbine: A LSTM and VMD‐based optimization approach

AbstractThe reversible pump‐turbine plays an important role in hydropower stations, but pressure pulsation during their operation affects their performance and lifespan. Accurate prediction of pressure pulsation signals can provide an important basis for energy planning and stable operation of pumped storage units, thereby promoting sustainable development of the environment. This study introduces an optimization method that combines long short‐term memory (LSTM) and variable mode decomposition (VMD) to enhance the prediction accuracy of pressure pulsation signals. First, by decomposing the pressure pulsation signal into multiple relatively stable subsequence components using VMD, the characteristics of the original signal become more distinct. Subsequently, individual LSTM‐based time series prediction models were constructed for each modal function, and the hyperparameters related to subsequence were optimized using the sparrow search algorithm. To validate the efficacy of the proposed approach, this paper conducted experiments using pressure pulsation signals of a pump‐turbine obtained through numerical simulation. The experimental data was divided into training and testing sets, with the former used to train the LSTM model and the latter used for validation. The experimental results show that the optimized VMD with an optimized LSTM method can effectively improve the prediction accuracy of pressure pulsation signals in reversible pump‐turbine.

The Short Time Prediction of the Dst Index Based on the Long-Short Time Memory and Empirical Mode Decomposition–Long-Short Time Memory Models

The Dst index is the geomagnetic storm index used to measure the energy level of geomagnetic storms, and the prediction of this index is of great significance for geomagnetic storm studies and solar activities. In contrast to traditional numerical modeling techniques, machine learning, which emerged decades ago based on rapidly developing computer hardware and software and artificial intelligence methods, has been unprecedentedly developed in geophysics, especially solar–terrestrial space physics. This study uses two machine learning models, the LSTM (Long-Short Time Memory, LSTM) and EMD-LSTM models (Empirical Mode Decomposition, EMD), to model and predict the Dst index. By building the Dst index data series from 2018 to 2023, two models were built to fit and predict the data. Firstly, we evaluated the influences of the learning rate and the amount of training data on the prediction accuracy of the LSTM model, and finally, 10−3 was thought to be the optimal learning rate. Secondly, the two models were used to predict the Dst index in the solar active and quiet periods, respectively, and the RMSE (Root Mean Square Error) of the LSTM model in the active period was 7.34 nT and the CC (correlation coefficient) was 0.96, and those of the quiet period were 2.64 nT and 0.97. The RMSE and CC of the EMD-LSTM model were 8.87 nT and 0.93 in the active period and 3.29 nT and 0.95 in the quiet period. Finally, the prediction accuracy of the LSTM model in the short time period was slightly better than the EMD-LSTM model. However, there will be a problem of prediction lag, which the EMD-LSTM model can solve and better predict the geomagnetic storm.

Antagonistic, Cosmopolitan and Agonistic Memory Modes in Postmemorial Third-Generation Holocaust Narratives of Lithuanian Diaspora Writers

This article examines the challenges of coming to terms with Holocaust perpetratorship as depicted in postmemorial third-generation Holocaust literature by Lithuanian diaspora writers Rita Gabis, Julija Šukys, and Silvia Foti. It also outlines the mnemonic strategies used to reconcile contradictory historical narratives from the perspectives of both the victims and the perpetrators. The analysis demonstrates the authors’ approaches to portraying victims and perpetrators, their choices in framing conflicting historical accounts, and their exploration of individual actions within the context of collective national identities. The main variations observed in the structuring of conflict stem from differences in writing styles, levels of personal connection to familial history, and collective experiences of suffering. These variations are also intertwined with the deliberate silence surrounding the individual-historical narrative which the collective voluntarily retreated into.

Extraction, chemical, and dual-functional memory and threshold resistive switching characteristics of Elaeodendron buchananii extract

This work presents a study on the electrical properties of plant materials, which is a fascinating area of research. These materials show promising potential in mitigating the exponential increase in electronic waste by enabling the development of eco-friendly electronic devices that can biodegrade. This research article showcases findings on the extraction, chemical analysis, and dual-functional memory and threshold resistive switching of extracts from the Elaeodendron buchananii plant. We report threshold switching with impressive ON/OFF ratio of >103, a retention time of ≥103, and endurance of 20 cycles, for the sandwich type device consisting of Elaeodendron buchananii plant extract-based film positioned between silver and indium-doped tin oxide electrodes. After modifying the experimental parameters, specifically by reducing the voltage scan increments to 1mV and increasing the compliance current to 150μA, transition from threshold switching to bipolar memory mode was detected. This particular memory mode exhibited an ON/OFF ratio of 10 and demonstrated very low switching voltages of −0.37V and +0.22V, although for only a limited number of write/erase cycles. The switching mechanism for both the threshold and memory mode can be attributed to the formation of conductive filaments caused by the migration of oxygen vacancies.

Study of magnetoelastic demagnetization of locally magnetized steel

To improve the accuracy of determining mechanical stresses in steel by the method of magnetoelastic demagnetization (magnetoelastic memory), the magnetoelastic sensitivity of the material to elastic stresses is additionally measured by applying or removing an additional load of a known value. We present the results of studying the magnetoelastic demagnetization of locally magnetized steel by varying stresses of the uniaxial tension or compression against a background of the base load. The magnetoelastic sensitivity of the steel in a loaded state to variable stresses has been assessed. It is shown that the determination of the magnetoelastic sensitivity of the steel to variable loads makes it possible to improve the accuracy of estimating base stresses in the steel structure using magnetoelastic methods. It is revealed that a decrease in the strength of the magnetic scattering field of the local residual magnetization (LRM) of the steel after variable loading (or unloading) exponentially depends on the magnitude of the base stresses. The possibility of controlling the uniaxial mechanical stresses in steel structures in the magnetoelastic memory mode which is based on the dependence of the strength of magnetic scattering field of the local residual magnetization of the steel on the uniaxial stresses is shown. A method for monitoring uniaxial stresses in the elements of steel structures by the method of magnetoelastic memory is proposed using the measured magnetoelastic sensitivity of the material. The results obtained can be used in the development of a method for monitoring the stress-strain state of steel structures during operation under the simultaneous effect of static and dynamic loads.

A predictive coding approach to psychedelic virtual-induced hallucinations and creative cognition in aging.

Recent research has investigated the potential of psychedelic substances in treating various neurological and psychiatric disorders. In particular, there has been a growth in studies concerning the intersection of psychedelics, Virtual Reality (VR), and Cognitive Flexibility (CF). Indeed, the use of immersive technology allows the simulation of the perceptual and cognitive effects of psychedelic substances without the potential risks associated with them. CF is strongly associated with creative cognition, a complex cognitive mechanism involved in creative thinking and associated with the prefrontal cortex and the neural networks supporting executive functions, memory, attention, and spontaneous modes of thought. The Bayesian brain approach, which is rooted in predictive coding, has emerged as a promising framework for understanding the effects of psychedelic hallucinations on cognitive functioning. Psychedelic substances may enhance creativity by inducing a state of CF, allowing for a wider range of associations and possibilities to be explored and increasing openness to experience. A decline in cognitive abilities, including creative processing and divergent thinking, is observed during the aging process. In particular, studies on Mild Cognitive Impairment (MCI) show poorer performance in executive functions, including CF. The present paper suggests that psychedelic hallucinations induced by VR may help optimize the balance between top-down expectations and bottom-up sensory information. Therefore, enhanced CF and creativity may be crucial during the aging process for maintaining cognitive functions and preventing pathological conditions.

A 6T-3M SOT-MRAM for in-memory computing with reconfigurable arithmetic operations

Traditional von Neumann architecture bottlenecks such as the “memory wall” limit artificial intelligence (AI) development, and in-memory computing (IMC) as a new computing architecture can solve the above problems. Spin-orbit-torque-magnetic random access memory (SOT-MRAM) has very good advantages in IMC architecture because of its good compatibility with CMOS, high tunneling magnetoresistance (TMR) ratio, and high energy efficiency. In this paper, we propose a 6T-3M-based MRAM-IMC architecture with reconfigurable memory mode and logical operation mode. The functionality of the proposed architecture is validated using the 28 nm process design kit and the SOT-MTJ model.