Abstract
The thesis focus is on employing the Perceiver architecture to improve classification performance on the Breast Histopathology Images dataset, which comprises 162 whole-slide images at 40x magnification and 277,524 patches of size 50x50 (198,738 negative IDC and 78,786 positive IDC). The research involved meticulous data preprocessing to optimize model accuracy, utilizing Kaggle's platform and libraries for model training and evaluation. Besides, this study implemented and compared several neural network architectures, including Perceiver, Densenet201, VGG-16, and EfficientNetB0, evaluating their performance based on metrics such as accuracy, loss, and training time.