Perovskite Neuromorphic Engine for Transformer Architectures
This paper introduces a perovskite‐based memristive computing unit that performs all necessary operations for Transformer artificial neural networks entirely in the analog domain. The system achieves comparable performance to GPU‐based systems, but with only 1.7% of the energy consumption, boosting power efficiency by 58 times.
Zhenye Zhan+5 more
wiley +1 more source
Deep Convolutional Framelets for Dose Reconstruction in Boron Neutron Capture Therapy with Compton Camera Detector. [PDF]
Didonna A+5 more
europepmc +1 more source
HiCGen introduces a hierarchical deep learning framework to predict genome organization across spatial scales using DNA sequences and genomic features. The model enables cross‐cell‐type predictions and in silico perturbation analysis, revealing correlations between loop domains and higher‐order structures.
Jiachen Wei, Yue Xue, Yi Qin Gao
wiley +1 more source
Systematic Evaluation of Atrous Spatial Pyramid Pooling in U-Net for Pore Segmentation in Plasma Electrolytic Oxidation Coatings. [PDF]
Chu CW, Lu CM, Yeung WK.
europepmc +1 more source
Convolution Neural Networks by Image Net
openaire +1 more source
A Wearable In‐Pad Diagnostic for the Detection of Disease Biomarkers in Menstruation Blood
MenstruAI integrates a paper‐based biosensor into sanitary pads to detect biomarkers in menstrual blood, enabling accessible, lab‐free diagnostics health. The platform supports early disease detection, especially in underserved communities, while challenging menstrual stigma and opening pathways for scalable, inclusive, and sustainable population ...
Lucas Dosnon+3 more
wiley +1 more source
Vehicle Target Detection of Autonomous Driving Vehicles in Foggy Environments Based on an Improved YOLOX Network. [PDF]
Liu Z, Zhang H, Lin L.
europepmc +1 more source
Memristive In‐Memory Object Detection with 128 Mb C‐Doped Ge2Sb2Te5 PCM Chip
A memristive in‐memory object detection system is presented for edge computing based on a 128 Mb phase change memory chip (40 nm, 99.99999 % yield) enabling in‐memory vector‐matrix multiplication and max computation. A novel mixed‐precision weight mapping reduces analog‐to‐digital‐converter energy by 22.3×.
Chenchen Xie+13 more
wiley +1 more source
Computer-aided diagnosis of Haematologic disorders detection based on spatial feature learning networks using blood cell images. [PDF]
Alsamri J+8 more
europepmc +1 more source