Back to home

Tag

neural networks

3 articles

Multi-fidelity models for composite mechanics
Research/May 5

Multi-fidelity models for composite mechanics

A review of co-kriging, deep Gaussian processes, and neural nets for cheaper composite-mechanics prediction.

Geometry Matters: Understanding Neural Networks Through Manifolds
Research/Mar 31

Geometry Matters: Understanding Neural Networks Through Manifolds

Researchers use Riemannian geometry to analyze the intrinsic structure of neural network representations, revealing patterns hidden by traditional similarity metrics.

IF4: Smarter 4-Bit Quantization That Adapts to Your Data
Research/Mar 31

IF4: Smarter 4-Bit Quantization That Adapts to Your Data

MIT researchers propose a hybrid data format that switches between floating-point and integer representations, improving accuracy in 4-bit neural network quantization.