Giorgio Mariani

Giorgio Mariani is a PostDoctoral Researcher at the University of Milano-Bicocca, having completed his Ph.D. in Computer Science at Sapienza, University of Rome under the supervision of Prof. Emanuele Rodolà. His work bridges theoretical machine learning with practical applications in audio processing and computer graphics.

His primary research focuses on:

• Generative models, particularly autoregressive and diffusion-based approaches
• Audio and music synthesis technologies
• Computer graphics applications for animation and gaming
• Adversarial vulnerabilities in geometric data

Mariani's publication trajectory shows significant progression from geometric deep learning toward advanced audio generation techniques. His recent work on music generation using diffusion models has been accepted at ICASSP 2024-2025, with 'Latent Autoregressive Source Separation' earning an oral presentation at ICLR 2024. His foundational work on 'Generating Adversarial Surfaces via Band-Limited Perturbations' (Computer Graphics Forum, 2020) established his expertise in 3D shape analysis security.

Notable achievements include:

• Oral presentation at ICLR 2024 (top 5% acceptance rate)
• Multiple paper acceptances at ICASSP 2024 and upcoming 2025 conference
• Research bridging audio processing, computer graphics, and security domains

As a postdoctoral researcher, Mariani continues to push boundaries in generative AI for audio applications while maintaining connections to his foundational work in 3D geometric data analysis, demonstrating exceptional interdisciplinary research capabilities.