IBM Develops Machine Learning Algorithm Fit for Quantum Computers

IBM System Q quantum computer Credit: IBMIBM System Q quantum computer Credit: IBM

IBM announced today that it has developed and tested quantum algorithms that have shown the potential to enable practical machine learning applications on future generations of quantum computers.

Machine Learning Accelerated by Quantum Computers

IBM has recently said that it expects to achieve a “quantum advantage” over classical computers in the coming years. The more powerful quantum computers can then be paired with machine learning algorithms that have been optimized for this type of system to achieve results that wouldn’t have been possible on any classical computer or supercomputer previously. 

IBM was able to use short-depth circuits to classify data in a way that also dealt with the expected decoherence (loss of state) in a quantum computer. The classification was done by a machine learning algorithm that was optimized for quantum computing. The classification showed no errors, even though IBM’s quantum computers experienced decoherence.

Complex Feature Mapping

According to IBM, both classical and quantum computing algorithms can break down a picture, for example, into its individual pixels and then place those pixels in a grid based on each pixel’s color value. This technique is called “feature mapping,” and the more precise this data can be classified, according to specific characteristics, the better the machine learning system will perform. The goal of matching quantum computers with machine learning is to create more sophisticated feature maps, which can allow the artificial intelligence (AI) system to identify patterns that could be invisible to classical computers.

The feature-mapping algorithms IBM developed for quantum computers was only tested on a simulation of a two-qubit quantum computer, but it still showed that there is a promising path forward for machine learning algorithms that run on quantum computers. IBM believes its machine learning algorithms could soon classify far more complex data sets than any classical computer could handle.

IBM’s algorithms demonstrating how entanglement can improve AI classification accuracy will be available as part of IBM’s Qiskit Aqua, an open-source library of quantum algorithms.