AAI_2025_Capstone_Chronicles_Combined

Breast Tumor Classification Using Quantum Neural Networks

7

taken there is to eliminate features that are closely correlated, such as perimeter mean and

perimeter worst in Figure 3.

Other Approaches With This Data Set

This breast cancer data set lists many different projects utilizing it (H, 2021). The 30

features in this data set are too large to be simulated, so feature selection is mandatory to fit

within the ~20 qubit simulation constraint. There is an exponential slow down as the number of

qubits increases, so that upper limit of ~20 can only be reached with a large computational cost.

The exploratory data analysis and feature selection others have performed provided direction

for this project that might not be otherwise apparent. For example, Figure 4 clearly shows the

relation between two key features.

Figure 4

Relation of mean area and radius as it applies to diagnosis, from (Karakanlı, 2025) .

At the time of this writing, Kaggle lists 458 code submissions for this data set (H, 2021).

Some of the most commented on include doing (classical) neural networks from scratch, using

260