Miguel L. Martins

What are good representations?

I am a final year Computer Science PhD. candidate at the University of Porto, Portugal.

I had the pleasure to work in many different fields, from network science, natural language processing, multifractal analysis, to biomedical signal and image processing.

Nowadays, I am mainly interested on a more principled understanding of the miracles we observe everyday with modern deep learning models.

I am currently focused on links between robustness, generalization, accuracy and identifiability in self-supervised machine learning.

I also like to build and break things, especially my neovim config.

📘 Extended bio

News

[03/Jul/2025] - New blogpost: CircorDigiscope dataset available at HuggingFace dataset hub!
[30/May/2025] - Our paper on understanding squeeze-and-excitation dynamics was accepet for publication at EUSIPCO 2025!
[04/Apr/2025] - Our work on deep learning for metaplasia detection resulted in two publications: EMBC 2025 and a Journal article at Endoscopy 2025.
[08/Mar/2025] - Another tutorial session on ENEEB 2025 on heart sound segmentation.
[29/Oct/2024] - Hands on session in heart sound segmentation.
[26-30/Aug/2024] - Attended EUSIPCO 2024 and presented our work on singularity strength recalibration of FCNs.
[06/Set/2023] - IEEE JBHI journal accepted on Markov-based Neural Networks for Heart Sound Segmentation

–>