publications

2026

1. arXiv

   Noise Scheduling as Information-Guided Allocation in Diffusion Training

   Raya, Gabriel, Nguyen, Bac, Batzolis, Georgios, Takida, Yuhta, Stancevic, Dejan, Murata, Naoki, Lai, Chieh-Hsin, Mitsufuji, Yuki, and Ambrogioni, Luca

   arXiv preprint arXiv:2602.18647 2026

   Abs arXiv PDF

   We study diffusion training noise schedules as an information-allocation problem and propose InfoNoise, a data-adaptive method that allocates training effort using entropy-reduction profiles estimated during training.

2025

1. arXiv

   Memorization to Generalization: Emergence of Diffusion Models from Associative Memory

   Pham, Binh, Raya, Gabriel, Negri, Matteo, Zaki, Mohammed J., Ambrogioni, Luca, and Krotov, Dmitry

   arXiv preprint arXiv:2505.21777 2025

   Abs arXiv PDF

   We connect diffusion models to associative memory and show how memorization gives way to generalization as training progresses, offering a theoretical perspective on the emergence of generative behavior.

2. ICLR

   Dynamic Negative Guidance of Diffusion Models

   Koulischer, Felix, Deleu, Johannes, Raya, Gabriel, Demeester, Thomas, and Ambrogioni, Luca

   International Conference on Learning Representations 2025

   Abs arXiv OpenReview PDF

   We introduce dynamic negative guidance for diffusion models, enabling more flexible and effective steering of the generative process.

3. arXiv

   Measuring Semantic Information Production in Generative Diffusion Models

   Handke, Florian, Koulischer, Félix, Raya, Gabriel, and Ambrogioni, Luca

   arXiv preprint arXiv:2506.10433 2025

   Abs PDF

   We propose measures of semantic information production during the diffusion generative process.

2024

1. NeurIPS

   Spontaneous symmetry breaking in generative diffusion models

   Raya, Gabriel, and Ambrogioni, Luca

   Advances in Neural Information Processing Systems 2024

   Abs arXiv NeurIPS PDF Blog Code Poster Slides

   Generative diffusion models have recently emerged as a leading approach for generating high-dimensional data. In this paper, we show that the dynamics of these models exhibit a spontaneous symmetry breaking that divides the generative dynamics into two distinct phases:1) A linear steady-state dynamics around a central fixed-point and 2) an attractor dynamics directed towards the data manifold. These two "phases" are separated by the change in stability of the central fixed-point, with the resulting window of instability being responsible for the diversity of the generated samples. Using both theoretical and empirical evidence, we show that an accurate simulation of the early dynamics does not significantly contribute to the final generation, since early fluctuations are reverted to the central fixed point. To leverage this insight, we propose a Gaussian late initialization scheme, which significantly improves model performance, achieving up to 3x FID improvements on fast samplers, while also increasing sample diversity (e.g., racial composition of generated CelebA images). Our work offers a new way to understand the generative dynamics of diffusion models that has the potential to bring about higher performance and less biased fast-samplers.

2. MIA

   Diffusion models for out-of-distribution detection in digital pathology

   Linmans, Jasper, Raya, Gabriel, Laak, Jeroen, and Litjens, Geert

   Medical Image Analysis 2024

   Abs Paper

   We applied diffusion models for unsupervised out-of-distribution (OOD) detection in digital pathology, significantly outperforming existing methods in the Camelyon16 challenge. These models effectively identify OOD data with high accuracy and alter anomalous input to appear more normal. Despite not reaching supervised learning’s performance, diffusion models show considerable potential for anomaly detection in medical imaging.

3. arXiv

   Losing dimensions: Geometric memorization in generative diffusion

   Achilli, Beatrice, Ventura, Enrico, Silvestri, Gianluigi, Pham, Bao, Raya, Gabriel, Krotov, Dmitry, Lucibello, Carlo, and Ambrogioni, Luca

   arXiv preprint arXiv:2410.08727 2024

   Abs PDF

   We study geometric aspects of memorization in generative diffusion models and how dimensional collapse relates to memorization of training data.

2021

1. SPIE

   Neural image compression for non-small cell lung cancer subtype classification in H&E stained whole-slide images

   Aswolinskiy, W., Tellez, D., Raya, G., Woude, Lieke, Looijen-Salamon, Monika, Laak, Jeroen, Grunberg, Katrien, and Ciompi, Francesco

   2021

   Abs HTML PDF

   Classification of non-small-cell lung cancer (NSCLC) into adenocarcinoma (LUAD) and squamous cell carcinoma (LUSC) is essential to select the appropriate treatment for the patient. However, most machine learning approaches require many detailed annotations of whole slide images. We propose to use Neural Image Compression (NIC), which requires only slide-level labels, to classify NSCLC. We show that NIC approaches state of the art performance on lung classification when trained on >2,000 slides from the TCGA and TCIA databases. The models reach AUCs in the range 0.84-0.98 on several internal and external cohorts.

2020

1. MSc. Thesis

   Unsupervised Out-of-distribution detection in digital pathology

   Raya, G.

   2020

   PDF Slides