Quantum simulation and machine learning process similarities

0
2523 views

There are strong similarities between typical machine learning training processes on the one hand, and, on the other hand, the state preparation, time evolution and property readout stages of quantum system simulations or chemical synthesis.
Who might be the most advanced groups actively working on quantum based generalised machine learning architectures?

Giovanni D
81 months ago

2 answers

0

This may not be specific enough for what you're looking for, but there is an area known as Quantum Machine Learning, which sits at the intersection of quantum physics and machine learning. Have a look at the wiki on the topic for a full breakdown. As I recently answered on Quora, here are some highlights:
When it comes to machine learning benefiting from quantum mechanics, we talk about so-called quantum machine learning algorithms. These look to leverage the advantages of quantum computation in order to achieve higher efficiency.
In the other direction, machine learning can be used to analyze quantum systems. Here, ML is used to process large amounts of data gathered from experiments in order to characterize some quantum system.
There is also the benefit that comes from comparing learning systems (via ML) with physical systems (via QM) in terms of methodological and structural similarities.
Quantum enhanced machine learning is still largely theoretical but it has been implimented on small-scale quantum devices.
There are a number of approaches for realizing “quantum enhanced machine learning.” They generally involve connecting what is measured from the physical system to the kind of outputs needed to support learning systems.
For example, the amplitutdes of a quantum state can be associated with the inputs and outputs of computations. This can lead to exponentially compact representations. Other methods use amplitude amplification for unstructured search tasks needed for algorithms like k-NN, or the quadratic speedup when training a perceptron.
Quantum enhancements have also been applied to reinforcement learning where a “quantum agent” interacts with a classical environment, accessing that environment via superpositions.
Machine learning can often benefit from an ability to sample from high-dimensional probably distributions. There has been work done on applying quantum annealing hardware to train deep neural networks.
There is also a focus on quantum learning theory which is the quantum analog to computational learning theory. In this case the learner is a quantum information processing device.
In 2013 Google and NASA launched the Quantum AI Lab using D-Wave’s adiabatic quantum computer.

Sean McClure, Ph.D.
81 months ago
Many Thanks for the comprehensive answer Sean: appreciated. NASA AMES together with Google and D-wave are certainly carving in-roads in this strategic area, alongside other teams such as Zhaokai Li's team at the University of Science and Technology in Hefei (https://arxiv.org/abs/1410.1054) and others. - Giovanni 81 months ago
0

Thank you Giovanni. Agreed.
I will definitely look into Li's work. Thanks for the reference.

Sean McClure, Ph.D.
81 months ago

Have some input?