Rigetti develops a new quantum algorithm to supercharge unsupervised Machine Learning

In order to update classical computing resources, and traditional learning techniques, Rigetti has hybridized these practices with quantum processing abilities. They have trained a 19-qubit gate model processor to solve a clustering problem. Clustering is a machine-learning technique used to organize data into similar groups and is a foundational challenge in unsupervised learning.

Their 19Q quantum computer, available through its cloud computing platform, Forest, uses a quantum approximate optimization algorithm. The Forest platform is used for controlling the quantum computer and accessing the data it generates. This novel algorithm is combined with a gradient-free Bayesian optimization to train the quantum machine. This hybridization relies on Bayesian optimization of classical parameters within the quantum circuit. It reaches an optimal solution in fewer steps than would otherwise be expected by drawing cluster assignments uniformly at random. The runtime for 55 Bayesian optimization steps with N = 2500 measurements per step is approximately 10 minutes. Rigetti’s algorithm was able to reach the optimum in fewer than 55 steps (only about 25% of runs did not reach the optimum within 55 steps).

This algorithm can also be applied to other combinatorial optimization problems such as image recognition and machine scheduling.

Rigetti’s demonstration uses the largest number of qubits as compared to any other algorithm in a gate-based quantum processor. The algorithm showed robustness to realistic noise.

The entire algorithm is implemented in Python, leveraging the pyQuil library for describing parameterized quantum circuits in the quantum instruction language Quil. The Bayesian optimizer is provided by the open source package BayesianOptimization, also written in Python.

The above demonstration is just a basic example of how quantum computers can help solve machine learning problems. Hybrid approaches like this one form the basis of valuable applications for the first quantum computers. However, beating the best classical benchmarks will require more qubits and better performance.

Apart from working on developing new algorithms for quantum computing in machine learning, Rigetti Computing builds hardware and software to store and process quantum information.

You can learn more about their research on their blog.