spiking-mode-neural-networks

What problem does it solve?

Traditional training of high-dimensional spiking-mode networks suffers from high computational cost due to dense weight matrices. By applying Hopfield-style decomposition, this skill reduces complexity and reveals low-dimensional attractor structures, enabling more efficient exploration of neuromorphic learning paradigms.

Core Features & Use Cases

• Low-rank weight reconstruction through Phi, Psi, and scores to cut training costs.
• Mode-space training that provides transparent interpretation of network dynamics.
• Educational and research use in neuromorphic computing and neural manifold analyses.

Quick Start

Train a spiking-mode neural network using Hopfield decomposition on a sample dataset.