Physical Neural Networks for Quantum Computing

In this work, we successfully demonstrated a Physical Neural Networks (PNN) capable of minimizing a given cost function by gradient descent, laying the groundwork for applying PNNs as a low-energy solution for optimizing the noise seen by a silicon spin qubit within a given Coulomb diamond. Excluding the input of a digital clock signal, this PNN solves the gradient descent problem using entirely analog computing methods. We demonstrated our PNN on a simulated noise map and found it reduced the energy required to reach the minimal noise point when tested against a comparable digital computing program.