Alek Fröhlich

Ph.D. Student,

CSML, Italian Institute of Technology (IIT),

Genoa, Italy.

I am a second-year ELLIS Ph.D. student at the Italian Institute of Technology under the supervision of Massimiliano Pontil (IIT & UCL), Karim Lounici (École Polytechnique), and Vladimir Kostic (IIT & Univ. of Novi Sad).

Previously, I was a research fellow at the University of São Paulo Medical School working with Daniel Tiezzi.

I am broadly interested on bridging machine learning and science, where data is often scarce, models are commonly misspecified, and confounding factors abound. On the theoretical side, my interests lie in causality, uncertainty quantification, and developing rigorous representation learning algorithms. On the applied side, I have been using machine learning and statistics to explore the vast amount of information contained in hematoxylin and eosin whole slide images and improving decision-making in clinical settings, particularly breast cancer care.

news

2026	I was awarded the Google PhD Fellowship 2026!
2026	I mentored the project “Learning and Testing Independence in Geometric Data via Generative Models” at MIT’s Summer Geometry Initiative (SGI), organized by Prof. Justin Solomon.
2026	I presented three papers at ICML 2026 in Seoul, South Korea.

Publications

2026

Toward Scalable and Valid Conditional Independence Testing with Spectral Representations

Alek Frohlich, Vladimir Kostic, Karim Lounici, and 3 more authors

2026

Bib HTML PDF

@misc{frohlich2025scalablevalidconditionalindependence,
  title = {Toward Scalable and Valid Conditional Independence Testing with Spectral Representations},
  author = {Frohlich, Alek and Kostic, Vladimir and Lounici, Karim and Perazzo, Daniel and Tiezzi, Daniel and Pontil, Massimiliano},
  year = {2026},
  eprint = {2512.19510},
  archiveprefix = {arXiv},
  primaryclass = {cs.LG},
}

Outcome-Aware Spectral Feature Learning for Instrumental Variable Regression

Dimitri Meunier, Jakub Wornbard, Vladimir R. Kostic, and 5 more authors

2026

Bib HTML PDF

@misc{meunier2025outcomeawarespectralfeaturelearning,
  title = {Outcome-Aware Spectral Feature Learning for Instrumental Variable Regression},
  author = {Meunier, Dimitri and Wornbard, Jakub and Kostic, Vladimir R. and Moulin, Antoine and Fröhlich, Alek and Lounici, Karim and Pontil, Massimiliano and Gretton, Arthur},
  year = {2026},
  eprint = {2512.00919},
  archiveprefix = {arXiv},
  primaryclass = {stat.ML},
}

Equivariant Representation Learning for Symmetry-Aware Inference with Guarantees

Daniel Ordoñez-Apraez, Vladimir Kostić, Alek Fröhlich, and 3 more authors

2026

Bib HTML PDF

@misc{ordoñezapraez2025equivariantrepresentationlearningsymmetryaware,
  title = {Equivariant Representation Learning for Symmetry-Aware Inference with Guarantees},
  author = {Ordoñez-Apraez, Daniel and Kostić, Vladimir and Fröhlich, Alek and Brandt, Vivien and Lounici, Karim and Pontil, Massimiliano},
  year = {2026},
  eprint = {2505.19809},
  archiveprefix = {arXiv},
  primaryclass = {cs.LG},
}

2025

PersonalizedUS: Interpretable Breast Cancer Risk Assessment with Local Coverage Uncertainty Quantification

Alek Fröhlich, Thiago Ramos, Gustavo Motta Cabello Dos Santos, and 3 more authors

Proceedings of the AAAI Conference on Artificial Intelligence, Apr 2025

Bib HTML PDF

@article{Frohlich2025,
  title = {PersonalizedUS: {I}nterpretable Breast Cancer Risk Assessment with Local Coverage Uncertainty Quantification},
  volume = {39},
  doi = {10.1609/aaai.v39i27.35017},
  number = {27},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  author = {Fröhlich, Alek and Ramos, Thiago and Santos, Gustavo Motta Cabello Dos and Buzatto, Isabela Panzeri Carlotti and Izbicki, Rafael and Tiezzi, Daniel Guimarães},
  year = {2025},
  month = apr,
  pages = {27998-28006},
}

2024

Machine learning can reliably predict malignancy of breast lesions based on clinical and ultrasonographic features

Isabela Carlotti Buzatto, Sarah Abud Recife, Licerio Miguel, and 7 more authors

Breast Cancer Research and Treatment, Jul 2024

Bib HTML PDF

@article{Buzatto2024,
  title = {Machine learning can reliably predict malignancy of breast lesions based on clinical and ultrasonographic features},
  author = {Buzatto, Isabela Carlotti and Recife, Sarah Abud and Miguel, Licerio and Onari, Nilton and Faim, Ana Luiza Peloso and Bonini, Ruth Morais and Silvestre, Liliane and Carlotti, Danilo Panzeri and Fr\"{o}hlich, Alek and Tiezzi, Daniel Guimar{\~{a}}es},
  journal = {Breast Cancer Research and Treatment},
  year = {2024},
  month = jul,
  publisher = {Springer Science and Business Media LLC},
}

2023

Computational pathology pipeline enables quantification of intratumor heterogeneity and tumor-infiltrating lymphocyte score

Daniel Tiezzi, Alek Fröhlich, Stefano Pagnotta, and 1 more author

In ESMO Immuno-Oncology Congress, Dec 2023

Bib HTML

@inproceedings{Alek2023,
  title = {Computational pathology pipeline enables quantification of intratumor heterogeneity and tumor-infiltrating lymphocyte score},
  author = {Tiezzi, Daniel and Fröhlich, Alek and Pagnotta, Stefano and Chahud, Fernando},
  booktitle = {ESMO Immuno-Oncology Congress},
  organization = {ESMO},
  year = {2023},
  month = dec,
}