Browse Teach Curate

Mamba State Space Model Architecture for Cryptocurrency Price Forecasting

A selective state-space (Mamba) neural architecture, CryptoMamba, built for financial time-series forecasting: stacked C-Blocks of normalized Mamba layers and MLPs feed a merge layer that predicts the next day's Bitcoin closing price. Students learn how input-dependent state-space dynamics capture long-range dependencies and regime shifts more efficiently than LSTMs or Transformers, and how forecasts are coupled to trading algorithms to measure real-world returns.

KITT Text

CRYPTOMAMBA: LEVERAGING STATE SPACE MODELS FOR ACCURATE BITCOIN PRICE PREDICTION Mohammad Shahab

CryptoMamba is a Mamba-based selective state-space model (SSM) architecture for forecasting Bitcoin prices from historical daily open, high, low, close, and volume data. State-space models maintain a…