Tag: MNIST

A new decentralized federated learning framework uses particle swarm optimization to dynamically select clients, reporting 76.2% accuracy on CIFAR-10 under ...

After removing hidden-neuron permutations and ReLU rescaling freedoms, independently trained networks fell into reproducible task-dependent weight clusters, with dataset-specific structure ...

A new federated framework called FedAE-KMeans trains autoencoders locally on each device and clusters compressed latent representations on a central ...

New research shows that deep belief networks trained entirely without labels spontaneously organize their internal representations by class, collapse toward ...

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 study systematically benchmarks five CNN compression techniques across GPUs, CPUs, and noisy data, finding that the best choice ...

Researchers in Florence have built generative diffusion models in which quantum coherence and the intrinsic noise of a real IBM ...

Researchers have created a flexible memristor based on hexagonal boron nitride whose synaptic strength is programmed directly by bending, mimicking ...

A chemically doped two-dimensional tungsten diselenide synapse that senses light, stores memory and powers a 92 percent accurate neural network ...