New MoFE framework uses Mixture-of-Experts and FNOs for crypto forecasting

Researchers have introduced MoFE, a novel deep learning framework designed for cryptocurrency forecasting. This framework integrates Fourier Neural Operators (FNOs) within a Mixture-of-Experts (MoE) architecture, aiming to capture the complex dynamics of non-stationary cryptocurrency markets. MoFE utilizes adaptive FNO and Convolution dual-domain experts to learn spectral trends and microstructures, with a dynamic gating mechanism for adaptive strategy switching across market regimes. Experiments on Bitcoin data from January 2020 to December 2025 show MoFE achieving state-of-the-art performance in both T+1 and T+5 forecasting horizons, significantly improving directional accuracy and information coefficient, and leading to robust risk-adjusted performance in simulated trading. AI

IMPACT This novel framework could improve the accuracy and profitability of cryptocurrency trading by better capturing market dynamics.

RANK_REASON The cluster contains a research paper detailing a novel deep learning framework for cryptocurrency forecasting. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• adaptive FNO
• Bitcoin
• cryptocurrency
• Fourier Neural Operators
• Information Coefficient
• Mixture-of-Experts
• Sharpe ratio

• paper
• model release

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