From Rattle to Roar: Optimizer Showdown for MambaStock on S&P 500

2025/07/09 by Chan, Alena, Garmonina, Maria
#Computational Finance (q-fin.CP) #FOS: Computer and information sciences #FOS: Economics and business #Machine Learning (cs.LG)

paper · doi:10.48550/arxiv.2508.04707

Abstract

We evaluate the performance of several optimizers on the task of forecasting S&P 500 Index returns with the MambaStock model. Among the most widely used algorithms, gradient-smoothing and adaptive-rate optimizers (for example, Adam and RMSProp) yield the lowest test errors. In contrast, the Lion optimizer offers notably faster training. To combine these advantages, we introduce a novel family of optimizers, Roaree, that dampens the oscillatory loss behavior often seen with Lion while preserving its training speed.

Citations

• MambaStock: Selective state space model for stock prediction
• Mamba: Linear-Time Sequence Modeling with Selective State Spaces
• Symbolic Discovery of Optimization Algorithms
• Efficiently Modeling Long Sequences with Structured State Spaces

Related

• MambaStock: Selective state space model for stock prediction
• Optimizing Time Series Forecasting: A Comparative Study of Adam and Nesterov Accelerated Gradient on LSTM and GRU networks Using Stock Market data
• S&P 500 Trend Prediction
• Adaptive Alpha Weighting with PPO: Enhancing Prompt-Based LLM-Generated Alphas in Quant Trading
• ms-Mamba: Multi-scale Mamba for Time-Series Forecasting