There are great hopes that the enormous processing effect of quantum computers one day will release exponential advances in artificial intelligence. AI systems thrive when the machine training algorithms used to train them are given massive amounts of data for ingestion, classification and analysis. The more precisely this data can be classified according to specific features or functions, the better AI will perform. Quantum computers are expected to play a crucial role in machine learning, including the crucial aspect of accessing more computationally complex functional areas ̵
In a new natural science document entitled "Monitored Learning with Quantum Enhanced Space Opportunities", we describe the development and testing of a quantum algorithm with the potential to enable machine learning on quantum computers in the near future. We have shown that as quantum computers become more powerful in the coming years and their Quantum Volume increases, they will be able to perform feature mapping, an important component of machine learning, on highly complex data structures on a scale far beyond reaching even the most powerful classic computers.
Our methods could also classify data using concise circuits that open a path to dealing with decoherence. Just as significantly, our functional mapping functioned as predicted: no classification error with our engineered data, even though the IBM Q systems' processors experienced decoherence.
Larger, better image
Function mapping is a way of separating data to access finer grain aspects of this data. Both classic and quantum learning algorithms can break down an image, for example by pixels, and place them in a grid based on the color value of each pixel. Then, the algorithms map individual data non-linearly to a high-dimensional space that breaks down the data according to the main functions. In the much larger quantum state room, we can separate aspects and functions of this data better than we could, in a functional overview created by a classic machine learning algorithm. Ultimately, the more precisely these data can be classified according to specific features or functions, the better AI will perform.
The goal is to use quantum computers to create new classifiers that generate more sophisticated data cards. Thus, researchers can develop more efficient AIs, which can, for example, identify patterns in data that are invisible to classic computers.
We have developed a blueprint with new quantum data classification algorithms and feature maps. It is important for AI because the bigger and more different a dataset is, the harder it is to separate these data into meaningful classes to train a machine learning algorithm. Poor classification results from the machine's learning process could introduce unwanted results; for example, impair a medical device's ability to identify cancer cells based on mammography data.
We found that even in the presence of noise, we could consistently classify our constructed data with perfect accuracy during our test. Today's quantum computers struggle to keep their qubits in quantum statistics for more than a few hundred microseconds, even in a highly controlled laboratory environment. This is important because qubits need to remain in this state for as long as possible to make calculations.
Our algorithms show how entanglement can improve AI classification accuracy, will be available as part of IBM's Qiskit Aqua, an open source quantum algorithm library that developers, researchers, and industry experts can use to access quantum computers via classic applications or common programming language such as Python.
We are still far from obtaining Quantum Advantage for machine learning – the point where quantum computers surpass classic computers in their ability to perform AI algorithms. Our research does not yet show Quantum Advantage because we minimized the scope of the problem based on our current hardware features, using only two qubits of quantum compatibility capabilities that can be simulated on a classic computer. But the feature mapping methods we promote can soon classify much more complex datasets than what a classic computer could do. What we have shown is a promising way forward.
Researchers successfully simulate a 64-qubit circuit
Vojtěch Havlíček et al. Supervised learning with quantum enhanced traction rooms, Nature (2019). DOI: 10,1038 / s41586-019-0980-2