Explainable machine learning on medical imaging
Machine learning and deep learning are techniques for recognizing patterns that can be applied to medical images (X-rays, CT scan, photos). However, the black-box nature of these approaches has restricted clinical use, so it is important to introduce methods that make these approaches interpretable or explainable.
Bachelor or master degree
Analysis of Deep learning architecture
In this thesis we address implementations and case studies, identifying best design practices and evaluating modern deep learning architectures such as Convolutional NN (VGG, ResNet, Inception, mask-rcnn, etc.), temporal NN (RNN, LSTM, GRU, Transformers, etc.), generative architecture (Autoencoder, GAN, etc.), Graph NN; providing benchmark performance on datasets from the literature or generated.
Bachelor or master degree
Structural health monitoring
Structural health monitoring aims to detect and identify any deviation from a baseline condition, typically a damage-free baseline, to track relative structural integrity. Starting from signals collected by sensors installed on the building, the main tasks of this domain are damage detection, damage localization and damage quantification, using machine learning and deep lerning techniques. This work can be made in collaboration with the Politecnico di Milano or with Move srl.
Bachelor or master degree