New PyTorch environment DFSC enhances fractional scientific machine learning

Researchers have developed DFSC, a PyTorch environment designed for fractional scientific machine learning. This system utilizes a Mittag-Leffler Spectral Layer (MLSL) to differentiate between known fractional propagation and data-driven corrections, allowing neural modules to learn only unresolved dynamics. DFSC optimizes fractional orders and residual-network parameters jointly, employing an adaptive algorithm to meet specified error tolerances. The system supports various operator paths, trainable fractional orders, and direct inverse problems, offering significant speedups on CPUs and GPUs when reusing prepared Lanczos bases. AI

IMPACT Introduces a novel framework for fractional scientific machine learning, potentially improving efficiency and accuracy in specialized AI applications.

RANK_REASON The cluster describes a new method and software environment for fractional scientific machine learning, detailed in an arXiv paper. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• DFSC
• Lanczos
• Mittag-Leffler Spectral Layer
• mlsl
• PyTorch
• RTX 5070

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →