Skip to content

Critical Reservoir Initialization Criterion

Home / Concepts / Critical Reservoir Initialization Criterion

Critical Reservoir Initialization Criterion

Auto-generated stub. Edit this file to add more details.

A random-matrix-based criterion for identifying the critical gain parameter that separates ordered from chaotic regimes in recurrent neural network initialization.

Why It Matters

Provides a principled, theoretically-grounded method for weight initialization in recurrent structures, bridging the gap between random matrix theory and empirical reservoir performance.

Evidence

we derive a simple criterion to estimate the critical g_c for a broad class of recurrent architectures and we show that it closely tracks the gain at which a gated-RNN reservoir achieves peak performance on a chaotic forecasting task.

Metadata & Links