The Key Differences Between Real and Complex-Valued State Space Models | HackerNoon
Briefly

Most prior State Space Models (SSMs) use complex numbers to optimize performance, but we find real-valued SSMs outperform them in some contexts, suggesting a continuous-discrete spectrum influence.
We propose using a simplified architecture for Selective State Space Models, emphasizing efficient implementation and demonstrating its effectiveness through empirical evaluations across various tasks, including language and DNA modeling.
Read at Hackernoon
[
|
]