HAN-Mamba model enhances financial volatility forecasting with Mamba architecture

Researchers have developed HAN-Mamba, a novel hybrid architecture that integrates selective state space models (Mamba) with a hierarchical structure for improved financial volatility forecasting. This model replaces the computationally intensive Transformer encoders of its predecessor, HAN-T, with Mamba encoders, resulting in a significant reduction in parameters and improved performance on the Optiver Realized Volatility Prediction benchmark. HAN-Mamba demonstrates enhanced accuracy by extending its high-frequency context processing and enabling constant-time streaming updates, outperforming its attention-based counterpart. AI

IMPACT This hybrid architecture could offer more efficient and accurate forecasting for time-series data in finance and other domains.

RANK_REASON The item is a research paper detailing a new model architecture. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• DagsHub
• Gotit.pub
• HAN-Mamba
• Hugging Face
• IArxiv
• Mamba
• Mihai Bogdan Deaconu
• Optiver Realized Volatility Prediction benchmark
• ScienceCast
• transformer

• paper
• model release

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