The Post-Connectionist Era Begins Here.

Spectral Morphogenetic Resonance Networks

Zero-Backprop O(|E|) Complexity Zero-Forgetting

A radical departure from Gradient Descent. MoReNet encodes information as differential wave equations propagating across non-Euclidean graph Laplacians. Learning is physical topological erosion.

Clone Repository Read Quickstart

Why Deep Learning is Stagnating

Thermodynamic Dead-End

Backpropagation requires massive global energy to compute partial derivatives. MoReNet uses purely local physical wave propagation.

Catastrophic Forgetting

Standard networks overwrite weights, destroying old memories. MoReNet carves distinct topological "canyons" on orthogonal harmonics.

Memory-Compute Coupling

Instead of separating compute and weights, MoReNet's topology is the memory, and wave propagation is the compute.

Spectral Cymatics

MoReNet replaces linear algebra layers with the Graph Wave Equation:

$$ \frac{\partial^2 \Psi}{\partial t^2} + c^2 \mathcal{L}\Psi = 0 $$

We bypass the expensive $ \mathcal{O}(|V|^3) $ eigenvalue decomposition using Chebyshev Polynomials scaled by Bessel functions of the first kind ($ J_n $), ensuring the semantic soliton focuses laser-sharp energy:

                $$ \cos(\sqrt{\lambda} \cdot t) \approx J_0(t) + 2 \sum_{n=1}^{K} (-1)^n J_{2n}(t) T_{2n}(\hat{\mathcal{L}}) $$
            

Topological Erosion

Learning is Hebbian plasticity driven by wave interference:

                $$ \frac{\partial W_{ij}}{\partial t} = \alpha \cdot \Theta(|\Psi_{tot,i}| \cdot |\Psi_{tot,j}| - \text{plasticity}) $$
            

The MoReNet Horizon: Targets & Applications

MoReNet is not just a software framework; it is an abstraction layer for the next generation of physical computing. By eliminating backpropagation, we unlock entirely new frontiers:

⚡

Photonic AI Chips

The ultimate target. Compiling MoReNet onto optical and acoustic resonators, where learning and wave propagation occur passively at the speed of light with near-zero thermodynamic cost.

🧠

Edge Robotics

Enabling robots to learn continuously in the real world. New skills carve new topological canyons without overwriting or destroying previously learned motor controls.

🔗

Relational Logic

Resolving complex logical reasoning natively through graph topology. Bypassing the quadratic bottlenecks of Transformer Attention with pure geometric wave mechanics.

Experience the Physics of Learning

The background of this website is a live simulation of a MoReNet topology. You are not just reading about it; you can touch it right now.

🖱️

1. Inject Energy

Move your cursor over the dark background or click to release a massive blast of raw wave energy into the graph nodes.

🌊

2. Wave Diffusion

Unlike Backpropagation, energy flows passively through the topology. Watch the neon ripples travel along the edges.

✨

3. Topological Erosion

When waves meet, their constructive interference permanently brightens the connections, carving memory directly into space.