arXiv Science⌕ Search
Home/Archive/arXiv

arXiv · 1902.06886

SPINBIS: Spintronics based Bayesian Inference System with Stochastic Computing

Xiaotao Jia·Jianlei Yang·Pengcheng Dai·Runze Liu·Yiran Chen·Weisheng Zhao
View original source ↗Read source document ↗Export metadata ↓

Abstract

Bayesian inference is an effective approach for solving statistical learning problems, especially with uncertainty and incompleteness. However, Bayesian inference is a computing-intensive task whose efficiency is physically limited by the bottlenecks of conventional computing platforms. In this work, a spintronics based stochastic computing approach is proposed for efficient Bayesian inference. The inherent stochastic switching behaviors of spintronic devices are exploited to build stochastic bitstream generator (SBG) for stochastic computing with hybrid CMOS/MTJ circuits design. Aiming to improve the inference efficiency, an SBG sharing strategy is leveraged to reduce the required SBG array scale by integrating a switch network between SBG array and stochastic computing logic. A device-to-architecture level framework is proposed to evaluate the performance of spintronics based Bayesian inference system (SPINBIS). Experimental results on data fusion applications have shown that SPINBIS could improve the energy efficiency about 12X than MTJ-based approach with 45% design area overhead and about 26X than FPGA-based approach.

Explore related subjects

cs.ETcs.AR

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Xiaotao Jia, Jianlei Yang, Pengcheng Dai, Runze Liu, Yiran Chen, Weisheng Zhao. 2019-02-19. SPINBIS: Spintronics based Bayesian Inference System with Stochastic Computing. https://doi.org/10.1109/tcad.2019.2897631

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

ASTRA: Toward Agentic AI for Intelligent Device-Network-Cloud Synergy in Next-Generation Mobile Communication

The evolution toward next-generation mobile communication systems demands intelligence-native networks capable of autonomously adapting to user intent, yet the prevailing 3GPP protocol-driven device-network-cloud (DNC) architecture imposes three structural bottlenecks: protocol-constrained decision spaces confining optimization to predefined parameter subsets, cascaded information asymmetry from lossy interface compression that strips semantic context and causes intent miscalibration, and inherently reactive coordination mechanisms that trigger actions only after performance degradation. This paper proposes an autonomous agentic AI paradigm named Agentic Synergy for Telecommunication Resource Autonomy (ASTRA), which introduces a three-tier agent layer, including device agent, network agent, and cloud agent, decoupling network intelligence from the underlying hardware infrastructure. These agents collaborate through bidirectional semantic channels, including semantic intent messages, capability abstraction messages, global directives, and peer coordination, executing a six-phase cycle of perceive, reason and predict, communicate, decide, act, and learn that transforms reactive protocol-driven operations into proactive, intent-calibrated optimization over the full decision space. Validated through system-level simulations in two representative scenarios, ASTRA achieves a 13.1\% average throughput gain in dense-crowd cell selection by redistributing UEs from congested cells via semantic load exchange, and an 18.2\% passive handover reduction in high-speed mobility through predictive trajectory-aware coordination, providing initial evidence that the proposed agentic framework accesses solution regions structurally inaccessible under protocol-constrained architectures.

cs.ET↗
arXiv2026

A Closed-Form Molecule-Release Rule for Diffusion-Based Molecular Communications with Ligand Receptors

The number of molecules released per bit is a fundamental design variable of diffusion-based molecular communication (MC), and ligand-receptor reception breaks the more-is-better intuition. Too few molecules leave the bound-receptor observations buried in binding noise, while too many amplify the accumulated intersymbol interference and saturate the finite receptor population, again making the observations indistinguishable. Reliability therefore peaks in an interior operating region whose location seems to require an exhaustive search over the channel dynamics. In this paper, we show that this search can be obviated for a biologically plausible receiver that compares consecutive bound-receptor counts without channel state information or a decision threshold. We derive a closed-form transmission rule, which sets the number of molecules released per bit such that the receptor dissociation constant equals the geometric mean of the two bit-conditioned received concentration levels, prove that it exactly minimizes the bit error probability of a memoryless binomial receptor model, and express it in the physical channel parameters through an Euler--Maclaurin evaluation of the interference. Time-domain Monte Carlo sweeps of the channel and receptor parameters, corroborated by particle-based simulations, show that the empirically optimal release count coincides with the prediction or lies above it by a small factor.

cs.ET↗
arXiv2026

Composability rather than computation sets the cost of an analog EML hardware fabric

The operator eml(x, y) = exp(x) - ln(y) with the constant 1 generates the elementary functions, a continuous counterpart to NAND. Whether it yields a useful fabric had not been asked of hardware. We ask in network models, circuit simulation and SkyWater 130 nm layout. Four bipolar junctions evaluate the operator for 13 fJ, beating a width-matched digital datapath by 4-134x. The fabric assembled from them is not cheap: it loses to resource-matched baselines, and over the reals its grammar excludes trigonometry. Amplifiers holding those junctions' operating points take 74.5% of a cell's current, so a cell costs 3000 times what they spend. Extracted non-idealities cost 2.6x when a cell must hold a value and nothing when it need only be repeatable. Sharing them across cells recovers two of the three orders. The premise was that a universal primitive licenses a uniform machine. It survives in the primitive and fails in the machine.

cs.ET↗