LLM-guided evolutionary program synthesis discovers new best-known 2D finite low-discrepancy sets (N≥40), improves 3D benchmarks beyond proven optimality, and evolves Sobol’ parameters that reduce rQMC error in 32-D option pricing.

We propose diffusion similarity networks (DSNs), which employ rotationally invariant spherical harmonic features derived from diffusion magnetic resonance imaging (dMRI), to map gray matter structural organization.

This paper presents neuromaps derived from diffusion tensor, diffusion kurtosis, mean apparent propagator, and neurite orientation dispersion and density imaging of the young adult cerebral cortex with four latent factors: diffusion kurtosis, simple diffusivity, complex diffusivity, and fractional anisotropy

We employ a SWIN model, trained on HCP data and conditioned on registered T1 scans, to perform generalized dMRI denoising with DTI requiring only 90 seconds of scan time.