Can we use the capabilities of quantum computing to enhance neural networks?

Question

Given that qubits can exist in a superposition, meaning a qubit can represent probabilities of every possible state, what if we leverage this idea in artificial neural networks? Specifically, in scenarios where we require a softmax layer for multi-class classification, could we use the power of qubits in the output layer to reduce the number of nodes required to measure the probabilities of a prediction belonging to a particular class?

Answer 1

There's already a discipline called quantum machine learning.

The paper Challenges and opportunities in quantum machine learning focuses on quantum neural networks and quantum deep learning (according to the abstract), but feel free to look for other resources.