# News in Quantum Machine Learning

Some weeks ago, we successfully trained a quantum machine learning model on IBM quantum computers. Our QML algorithm reached the accuracy level of classical ML in solving a simple problem, the ternary classification of the Iris flower dataset. This well known dataset is typically used as a test case for statistical learning techniques.

Ref. Polyadic Quantum Classifier — arXiv:2007.14044

Today, we release our code to contribute to the collective quest towards QML, encouraging teams around the world to try their own datasets, improve the existing algorithms and come up with new ones.

# How does this result stand in the current QML landscape ?

I am not a scientific journalist so I prefer to rely on the excellent piece of Sophia Chen where she interviews prominent researchers in the field. To say the least their assessments are not washed with enthusiasm:

- “
*There is a lot more work that needs to be done before claiming quantum machine learning will actually work*” - “
*I have not seen a single piece of evidence that there exists a meaningful [machine learning] task for which it would make sense to use a quantum computer and not a classical computer*”

Well… regarding the first statement, our algorithm shows now empirical evidence that QML does indeed work, at least for some problems. And, as we keep running more experiments, the set of problems for which QML work can only increase.

The second statement, about the possible quantum advantage of QML, has yet no answer. However, we can note that for claiming quantum advantage we need:

- To have a quantum algorithm that performs better than its classical counterpart.
- To be able to verify that the result of the algorithm is correct

Hence, in the case of machine learning, to prove quantum advantage, it is a requisite that the algorithm did learn from some data correctly.

This sketches a roadmap for QML: try increasingly complex learning tasks and improve how it is done. If a quantum algorithm works successfully, for some task, for some data, and performs better than the best classical one, for some criteria, we can claim quantum advantage.

So let’s do it, let’s try things.

— Joaquín Keller, Head of Quantum Machine Learning, Entropica Labs