Results 91 to 100 of about 2,053,524 (303)

Single‐Step Analog In‐Memory Matrix Computation in Three‐Dimensional Circuits

Advanced Intelligent Systems, EarlyView.
A wooden model illustrates a 3D memristive array designed for single‐step kernel convolutions. It shows how two‐dimensional input data are mapped into the 3D array and convolved in parallel with stored kernels. Color‐coded connection points represent programmed memory‐element values, with each color corresponding to a specific kernel weight.
Alireza Jaberi Rad, Koushik Kumar, Fatemeh Kiani, Yi Huang, Ali Abdel‐Maksoud, Zheshun Xiong, Yunzhi Ling, Wuyu Zhao, Qiangfei Xia   +8 more
wiley   +1 more source

Exploiting nested task-parallelism in the H-LU factorization [PDF]

, 2019
We address the parallelization of the LU factorization of hierarchical matrices (-matrices) arising from boundary element methods. Our approach exploits task-parallelism via the OmpSs programming model and runtime, which discovers the data-flow ...
Börm, Steffen, Aliaga, José I., Christophersen, Sven, Aliaga Estellés, José Ignacio, Carratalá-Sáez, Rocío, Beltran, Vicenç, Beltran Querol, Vicenç, Quintana-Orti, Enrique S.   +7 more
core   +1 more source

Towards Task-Parallel Reductions in OpenMP

, 2015
Reductions represent a common algorithmic pattern in many scientific applications. OpenMP* has always supported them on parallel and worksharing constructs. OpenMP 3.0’s tasking constructs enable new parallelization opportunities through the annotation of irregular algorithms.
Jan Ciesko, Sergi Mateo, Xavier Teruel, Xavier Martorell, Eduard Ayguadé, Jesús Labarta, Alex Duran, Bronis R. de Supinski, Stephen Olivier, Kelvin Li, Alexandre E. Eichenberger   +10 more
openaire   +3 more sources

A Scalable and Resource‐Efficient Pipelined p‐Computer for Probabilistic Ising Machines

Advanced Intelligent Systems, EarlyView.
(a) Block diagram of the portfolio optimization problem: given M assets, the goal is to determine the optimal weights w that maximize the expected return (based on the mean historical assets return u), while minimizing the risk, quantified by the assets covariance matrix S.
Deborah Volpe, Eleonora Raimondo, Andrea Grimaldi, Pedram Khalili Amiri, Stefano Chiappini, Anna Giordano, Mario Carpentieri, Hyunsoo Yang, Massimo Chiappini, Giovanni Finocchio   +9 more
wiley   +1 more source

Vertical NAND Flash Memory‐Based Bayesian Neural Networks With Intrinsic Gaussian Weight Sampling

Advanced Intelligent Systems, EarlyView.
Commercial vertical NAND flash memory supports Bayesian neural networks by using paired cells to produce bell‐shaped weight distributions. Read voltage controls their spread without additional memory cells or changes to the array structure. Simulations based on device measurements demonstrate improved uncertainty estimation and accuracy gains of up to ...
Sung‐Ho Park, Mingyun Oh, Dongbeen Shin, Jaehyeon Kim, Jonghyun Ko, Jiseong Im, Yeongheon Yang, Gyuweon Jung, Jong‐Ho Lee   +8 more
wiley   +1 more source

Scalable deep text comprehension for Cancer surveillance on high-performance computing

BMC Bioinformatics, 2018
Background Deep Learning (DL) has advanced the state-of-the-art capabilities in bioinformatics applications which has resulted in trends of increasingly sophisticated and computationally demanding models trained by larger and larger data sets.
John X. Qiu, Hong-Jun Yoon, Kshitij Srivastava, Thomas P. Watson, J. Blair Christian, Arvind Ramanathan, Xiao C. Wu, Paul A. Fearn, Georgia D. Tourassi   +8 more
doaj   +1 more source

Modeling the Benefits of Mixed Data and Task Parallelism [PDF]

, 1995
Mixed task and data parallelism exists naturally in many applications, but utilizing it may require sophisticated scheduling algorithms and software support.
James Demmel, Katherine Yelick, Chakrabarti, Soumen, Demmel, James, Yelick, Katherine, Soumen Chakrabarti   +5 more
core   +1 more source

Kinematic‐Aware Diffusion Model Correction for Human Motion Generation via Pose Transfer

Advanced Intelligent Systems, EarlyView.
Overall architecture of the proposed video kinematic pose correction pipeline. The framework integrates diffusion‐based video generation, multi‐estimator skeleton extraction, temporal Kalman smoothing, biomechanical constraint enforcement, and pose‐conditioned frame regeneration via Pose‐guided Image Synthesis and Editing. This paper presents Kinematic‐
Alaa Abdullah Albaghdadi, Ahmad R. Naghsh‐Nilchi   +1 more
wiley   +1 more source

User privacy prevention model using supervised federated learning‐based block chain approach for internet of Medical Things

CAAI Transactions on Intelligence Technology, EarlyView., 2023
Abstract This research focuses on addressing the privacy issues in healthcare advancement monitoring with the rapid establishment of the decentralised communication system in the Internet of Medical Things (IoMT). An integrated blockchain homomorphic encryption standard with an in‐build supervised learning‐based smart contract is designed to improvise ...
Chandramohan Dhasarathan, Mohammad Kamrul Hasan, Shayla Islam, Salwani Abdullah, Shailesh Khapre, Dalbir Singh, Abdulaziz A. Alsulami, Ali Alqahtani   +7 more
wiley   +1 more source

GCN and DRL based on dependent task offloading mechanism in edge computing

Digital Communications and Networks
Task offloading is critical for optimizing resource allocation in edge computing systems. In practical scenarios, user applications often comprise multiple interdependent tasks, where both task dependencies and parallelism strongly affect offloading ...
Ruiqi Tong, Shaoyong Guo, Xuesong Qiu, Feng Qi, Dongxiao Yu   +4 more
doaj   +1 more source