Quantum ML & GPU Acceleration
PHASE 9A state-vector variational quantum classifier trained with parameter-shift gradients in a background worker, plus a WebGPU compute path that falls back to the CPU automatically.
Problem
Circuit
2 qubits · 8 trainable parameters
Optimiser
Gradients are exact, from the parameter-shift rule — every parameter costs two extra circuit evaluations per sample.
Status
120 samples · 84 train / 36 test
epoch 0 / 20
GPU: …
Loss and accuracy by epoch