Can quantum machine learning predict shipments that will be delivered late?
Tees Valley-based data science and AI development company Seer focuses on the maritime sector. Its Flytta product automates the creation of a large number of customs declarations quickly, from structured and semi-structured data. The Seer team’s challenge during QTAP was aimed at identifying shipments that will be delivered late.
What was done?
A PyTorch model of a quantum neural network (QNN) which combined classical machine learning components with the ORCA PT-1 quantum photonic boson sampler (quantum computer) was designed and built to identify late packages from an example data set. During the experiment, training data from Amazon about package shipments including whether the package was delivered late were loaded into the QNN.
What were the results?
It was not possible to train the QNN in this experiment. Quantum computing is a new technology and it’s not possible to always predict when it will be successful. In this case, whilst the experiment was unsuccessful, the Seer team have developed good ideas about how to make the QNN easier to train, and appreciated the opportunity to learn more about quantum computing, as well as getting the chance to run experiments on the ORCA Computing Software Development Kit (SDK).
This case study highlights that even when quantum technology cannot yet achieve its goals, the process itself offers crucial learning and developmental opportunities for future success.
