The Resonant Phase Computer (RPC) is a hybrid analog/digital accelerator that solves combinatorial optimization by encoding variables as phases of nonlinear oscillators and constraints as couplings. The system evolves under controlled schedules to low-energy states corresponding to high-quality solutions.
Boundary Phase Resonance describes resonant phase patterns that emerge at interfaces between synchronized and unsynchronized regions. These patterns propagate information globally, helping the system avoid shallow minima and converge faster to stable solutions.
We target binary optimization expressed as Ising or QUBO forms. A typical QUBO objective seeks to minimize E(x) = x^T Q x subject to x in [0,1]^n. The equivalent Ising form uses spins s in [-1,+1]^n with energy E(s) = - sum over i of h_i s_i minus sum over i<j of J_ij s_i s_j.
Category | Spec | Notes |
---|---|---|
Compute core | Oscillator network (analog) | Phase‑based variable representation |
Control | FPGA + MCU | Timing, schedules, calibration |
Host link | PCIe Gen3 x4 (MVP) | Gen4 roadmap |
Board power | 50-150 W (target) | Workload dependent |
Problem size | 64 variables (MVP) | Scaling with ASIC roadmap |
SDK | Python, C++ | CLI utilities included |
Environment | Room temperature | No cryogenics required |
Form factor | PCIe add-in card | Rack systems for multi‑board |