Results 61 to 70 of about 142,362 (315)

A new approach to training neural networks using natural gradient descent with momentum based on Dirichlet distributions

Компьютерная оптика, 2023
In this paper, we propose a natural gradient descent algorithm with momentum based on Dirichlet distributions to speed up the training of neural networks. This approach takes into account not only the direction of the gradients, but also the convexity of
R.I. Abdulkadirov, P.A. Lyakhov
doaj   +1 more source

Engineering Deformation and Failure in Diamond Triply Periodic Minimal Surface Lattices via 3D Wall‐Thickness Grading

Advanced Engineering Materials, EarlyView.
The work demonstrates that strategic wall‐thickness grading in diamond triply periodic minimal surface lattices enables precise tuning of deformation and failure behavior under compression. Different gradation patterns guide how and where the structure collapses, improving energy absorption or promoting controlled brittle failure.
Giovanni Rizza, Danilo Bruson, Manuela Galati, Luca Iuliano   +3 more
wiley   +1 more source

Fractional Skyrmion Tubes in Chiral‐Interfaced 3D Magnetic Nanowires

Advanced Functional Materials, EarlyView.
In chiral 3D helical magnetic nanowires, the coupling between the geometric and magnetic chirality provides a way to create topological spin states like vortex tubes. Here, it is demonstrated how the breaking of this coupling in interfaced 3D nanowires of opposite chirality leads to even more complex topological spin states, such as fractional ...
John Fullerton, Naëmi Leo, Jakub Jurczyk, Claire Donnelly, Dédalo Sanz‐Hernández, Luka Skoric, Nicholas Mille, Stefan Stanescu, Donald A MacLaren, Rachid Belkhou, Aurelio Hierro‐Rodriguez, Amalio Fernández‐Pacheco   +11 more
wiley   +1 more source

Optimising continuous microstructures: a comparison of gradient-based and stochastic methods [PDF]

, 2004
This work compares the use of a deterministic gradient based search with a stochastic genetic algorithm to optimise the geometry of a space frame structure.
Hanna, S., Haroun Mahdavi, S.
core  

All‐in‐One Analog AI Hardware: On‐Chip Training and Inference with Conductive‐Metal‐Oxide/HfOx ReRAM Devices

Advanced Functional Materials, EarlyView.
An all‐in‐one analog AI accelerator is presented, enabling on‐chip training, weight retention, and long‐term inference acceleration. It leverages a BEOL‐integrated CMO/HfOx ReRAM array with low‐voltage operation (<1.5 V), multi‐bit capability over 32 states, low programming noise (10 nS), and near‐ideal weight transfer.
Donato Francesco Falcone, Victoria Clerico, Wooseok Choi, Tommaso Stecconi, Folkert Horst, Laura Bégon‐Lours, Matteo Galetta, Antonio La Porta, Nikhil Garg, Fabien Alibart, Bert Jan Offrein, Valeria Bragaglia   +11 more
wiley   +1 more source

Fast Ridge Regression with Randomized Principal Component Analysis and Gradient Descent [PDF]

, 2014
We propose a new two stage algorithm LING for large scale regression problems. LING has the same risk as the well known Ridge Regression under the fixed design setting and can be computed much faster. Our experiments have shown that LING performs well in
Foster, Dean P., Lu, Yichao
core  

Prospects of Electric Field Control in Perpendicular Magnetic Tunnel Junctions and Emerging 2D Spintronics for Ultralow Energy Memory and Logic Devices

Advanced Functional Materials, EarlyView.
Electric control of magnetic tunnel junctions offers a path to drastically reduce the energy requirements of the device. Electric field control of magnetization can be realized in a multitude of ways. These mechanisms can be integrated into existing spintronic devices to further reduce the operational energy.
Will Echtenkamp, Brahmdutta Dixit, Yifei Yang, Deyuan Lyu, Thomas Peterson, Qi Jia, Yu‐Chia Chen, Jian‐Ping Wang   +7 more
wiley   +1 more source

Balancing Privacy and Utility in Artificial Intelligence-Based Clinical Decision Support: Empirical Evaluation Using De-Identified Electronic Health Record Data

Applied Sciences
The secondary use of electronic health records is essential for developing artificial intelligence-based clinical decision support systems. However, even after direct identifiers are removed, de-identified electronic health records remain vulnerable to ...
Jungwoo Lee, Kyu Hee Lee
doaj   +1 more source

Peptide Sequencing With Single Acid Resolution Using a Sub‐Nanometer Diameter Pore

Advanced Functional Materials, EarlyView.
To sequence a single molecule of Aβ1−42–sodium dodecyl sulfate (SDS), the aggregate is forced through a sub‐nanopore 0.4 nm in diameter spanning a 4.0 nm thick membrane. The figure is a visual molecular dynamics (VMD) snapshot depicting the translocation of Aβ1−42–SDS through the pore; only the peptide, the SDS, the Na+ (yellow/green) and Cl− (cyan ...
Apurba Paul, Archith Rayabharam, Lisa Almonte, Eveline Rigo, Joshy Joseph, Ashutosh Kumar, Zhuxin Dong, Narayana Aluru, Gregory Timp   +8 more
wiley   +1 more source

ADADELTA: An Adaptive Learning Rate Method [PDF]

, 2012
We present a novel per-dimension learning rate method for gradient descent called ADADELTA. The method dynamically adapts over time using only first order information and has minimal computational overhead beyond vanilla stochastic gradient descent.
Zeiler, Matthew D.
core