Results 51 to 60 of about 2,749 (228)

Stochastic Rounding for LLM Training: Theory and Practice

CoRR
AISTATS ...
Kaan Ozkara, Tao Yu, Youngsuk Park
openaire   +3 more sources

A Simple and Efficient Stochastic Rounding Method for Training Neural Networks in Low Precision

CoRR, 2021
Conventional stochastic rounding (CSR) is widely employed in the training of neural networks (NNs), showing promising training results even in low-precision computations. We introduce an improved stochastic rounding method, that is simple and efficient.
Lu Xia, Martijn Anthonissen, Michiel E. Hochstenbach, Barry Koren   +3 more
openaire   +2 more sources

Room‐Temperature Skyrmionic Synapse in 2D Ferromagnet Fe3GaTe2 Operating via Collective Spin Texture Transformation

Advanced Materials, EarlyView.
We demonstrate a neuromorphic synapse in 2D Fe3GaTe2 flakes. The device operates via a current‐driven transformation from a skyrmion‐lattice to a stripe‐domain state, yielding a linear anomalous Hall resistance response with a tunable slope to enable multiply‐accumulate operations. Simulations confirm its viability in artificial neural networks.
Jixiang Huang, Tongji Zhu, PeiYu Cai, Xiaoming Ma, Zhen Wang, Ruifu Zhang, Yi Hao, Jingdi Lu, Liubing He, Jun Xu, Yuejie Zhang, Wanjun Jiang, Jing Tao, Shiming Lei, Pavel Parchinskiy, Jing Teng, Lingfei Wang, Wei Niu, Elton J.G. Santos, Xiaoqian Zhang, Peng Li   +20 more
wiley   +1 more source

Elephant‐Skin‐Inspired Porous Cementitious Tiles with Programmable Crack Networks for Passive Cooling

Advanced Materials, EarlyView.
Elephant‐skin‐inspired crack networks are programmed in porous diatomaceous earth (DE)‐cement composites using substrate‐guided, stress‐concentration induced fracture. The resulting crack lattices act as capillary conduits that redistribute water, while the porous matrix stores moisture.
Qingya Huang, Kun‐Hao Yu, Ji Yoon Bae, Yunchan Lee, Dorit Aviv, Shu Yang   +5 more
wiley   +1 more source

Stochastic rounding for memory-efficient digital simulation of synaptic plasticity using 8-bit floating-point

Neuromorphic Computing and Engineering
Simulating brain-scale networks digitally is often hindered by extensive memory access. In this context, using low-precision data types and more efficient models to represent state variables is a viable alternative to improve the scalability of the ...
Pablo Urbizagastegui, André van Schaik, Runchun Wang   +2 more
doaj   +1 more source

Resolving Heterogeneity of Targeted Lipid Nanoparticles Through Solution‐Based Biophysical Analyses

Advanced Materials, EarlyView.
AF4‐UV‐DLS‐MALS‐SAXS resolves previously inaccessible targeted lipid nanoparticle (tLNP) subpopulations that differ in size, shape, and composition. Correlation of subpopulation‐resolved biophysical properties with in vivo RNA delivery reveals that targeted placental transfection is associated with distinct tLNP subpopulations rather than ensemble ...
Hannah C. Geisler, Hannah C. Safford, Ajay S. Thatte, Marshall S. Padilla, Elisa Battistini, Hannah M. Yamagata, Violet M. Ullman, Alex Chan, Benjamin E. Nachod, Anushka Agrawal, Maxwell B. Watkins, Jesse B. Hopkins, Andrew Tsourkas, Kushol Gupta, Michael J. Mitchell   +14 more
wiley   +1 more source

Dynamic Mechanical Loading Modulates Extracellular Vesicle Production and Cargo in Meniscus Fibrochondrocytes

Advanced Materials Interfaces, EarlyView.
Dynamic mechanical loading of outer meniscus fibrochondrocytes increases extracellular vesicle (EV) yield and alters EV composition. When applied to progenitor cells, these mechanically primed EVs modulate the expression of matrix‐related genes.
Catherine Cheung, Zizhao Li, Sung‐Han Jo, Se‐Hwan Lee, Il Keun Kwon, Jina Ko, Su Chin Heo   +6 more
wiley   +1 more source

Multi-Objective Optimization of Cost and Resilience in water Distribution Networks Using MOGWO and EPANET [PDF]

منابع آب و تغییر اقلیم
Objective: Urban water distribution network (WDN) design is a critical component of water resources engineering, required to withstand challenges posed by population growth, rising demand, climate variability, and financial constraints.
Parisa-Sadat Ashofteh
doaj   +1 more source

Evaluation of stochastic particle dispersion modeling in turbulent round jets [PDF]

International Journal of Multiphase Flow, 2017
Abstract ODT (one-dimensional turbulence) simulations of particle-carrier gas interactions are performed in the jet flow configuration. Particles with different diameters are injected onto the centerline of a turbulent air jet. The particles are passive and do not impact the fluid phase.
Guangyuan Sun, John C. Hewson, David O. Lignell   +2 more
openaire   +1 more source

Benchmarking Coaxial and Angular Optical Emission Spectroscopy With Recommendations for Reliable Compositional In Situ Monitoring During Laser Powder Bed Fusion

Advanced Materials Technologies, EarlyView.
ABSTRACT Real‐time insight into local chemistry is critical for reliable part quality in additive manufacturing, especially laser powder bed fusion (PBF‑LB/M), where rapid thermal cycles and localized evaporation can undermine part performance. Optical emission spectroscopy (OES) offers non‑intrusive, in situ plume monitoring, but detection geometry ...
Philipp Gabriel, Bastian Jülich, Anna Rosa Ziefuss, Florian Eibl, Stephan Barcikowski   +4 more
wiley   +1 more source