View our Press Release

CUSP — Simulation of a Quantum AutoEnconder

What is CUSP?

Compressed Unsupervised State Preparation (CUSP) is a method for building more efficient quantum circuits by using a quantum autoencoder. The protocol performs a kind of circuit synthesis that, if training is successful, results in a more compact circuit. Since typically shorter-depth circuits are less prone to noise on a real quantum computer, this tool gives the opportunity to make a more accurate state preparation, resulting in better accuracy for a quantum computation. In this demo, we will use the example of improving a circuit which computes the ground state energies of molecular hydrogen at various bond lengths.

About CUSP

Zapata designed, from the ground up, CUSP — a quantum machine learning algorithm for building more efficient quantum circuits — and implemented it for Google’s quantum computers. CUSP draws upon the research by co-founder Alán Aspuru-Guzik, a pioneer in quantum simulation for chemistry and materials. CUSP is an important tool when a quantum algorithm is too large to optimize by hand as it can automatically compress a computation to fit on near term quantum computers. It is one of the many powerful, hardware-agnostic algorithms that Zapata is currently developing for Fortune 1000 companies in markets such as finance, pharmaceuticals, and materials. Zapata’s algorithms can run on the latest quantum hardware made by Google and other companies in this field. The USP algorithm and others like it will hasten breakthroughs and enable the next generation of discoveries in chemistry, materials, and artificial intelligence.