Trainable Reference Vectors Accelerate Neural Network Early Exits With Minimal Overhead
Researchers demonstrate that replacing BranchyNet classifier heads with trainable class reference vectors accelerates ResNet-50 inference by up to 1.67 times ...
Researchers demonstrate that replacing BranchyNet classifier heads with trainable class reference vectors accelerates ResNet-50 inference by up to 1.67 times ...
A federated learning framework combining feature distillation and momentum aggregation reduced communication volume by up to 97.6 percent while improving ...
Researchers have developed ShieldFL, an aggregation and blockchain auditing framework designed to counter Byzantine model-poisoning attacks in heterogeneous federated learning.
A comprehensive survey charts a decade of generative adversarial network developments, linking mathematical divergence choices to architecture search and practical ...
Researchers have developed a mixed-timestep spiking neural network conversion framework that matches layer-wise quantization to temporal steps, cutting inference latency ...
Researchers introduced Cross-Distribution Semantic Transformation to enhance adversarial transferability across neural network architectures without corrupting key features or imposing excessive ...
A simulated framework that dynamically coordinates stage allocation, federated aggregation, and runtime compression reduces edge artificial intelligence inference latency by ...
Researchers have developed Fancy GNG, an image augmentation method that partitions color distributions to improve neural network classification on color-sensitive ...
A new study shows federated learning's average accuracy can mask severe failures for individual clients, and that server-side re-weighting destabilizes ...
A new decentralized federated learning framework uses particle swarm optimization to dynamically select clients, reporting 76.2% accuracy on CIFAR-10 under ...
A new personalized federated learning framework called FedACC dynamically separates shared and personalized model parameters and clusters clients, reporting the ...
A six-client CIFAR-10 experiment found that layered federated-learning defenses improved accuracy and resilience to poisoning and adversarial attacks, while privacy ...
A knowledge-distilled pyramid transformer with reduced attention achieved F1-scores of 90.92%, 97.76% and 94.59% on three 32 × 32-pixel image-classification ...
A new federated framework called FedAE-KMeans trains autoencoders locally on each device and clusters compressed latent representations on a central ...
The COSMOS framework trains orthonormal subspace frames with conjugation-based gauge-style updates, holding geometric integrity at machine precision and matching a ...
A new computational neuroscience study shows that biologically constrained Hebbian learning can build useful visual representations without exact weight sharing, ...
Researchers have reimagined pixels as three-dimensional cuboids whose sizes are set by RGB values, using the empty space between them ...
A new federated learning framework called FedGPA balances global generalization and client personalization by aggregating neural network layers at different ...
A new Predictive Uncertainty Curve framework turns how AI models wobble under perturbations into a powerful, shadow-model-free membership inference attack ...
Researchers in China have developed a deep neural network fingerprinting method that explicitly separates stolen model copies from independently trained ...
We bring you the latest biotechnology news from best research centers and universities around the world. Check our website.
Bioengineer.org © Copyright 2023 All Rights Reserved.
Bioengineer.org © Copyright 2023 All Rights Reserved.