Marco Mondelli

Full Professor@Institute of Science and Technology Austria·奥地利·OpenReview

研究方向

information theory · learning theory · wireless communications · machine learning · data science · coding theory · theory of deep learning · high dimensional statistics · deep learning theory

Marco Mondelli

A Law of Data Reconstruction for Random Features (And Beyond)

High-dimensional Analysis of Synthetic Data Selection

Optimal Regularization for Performative Learning

Neural Collapse is Globally Optimal in Deep Regularized ResNets and Transformers

Wide Neural Networks Trained with Weight Decay Provably Exhibit Neural Collapse

High-dimensional Analysis of Knowledge Distillation: Weak-to-Strong Generalization and Scaling Laws

Attention with Trained Embeddings Provably Selects Important Tokens

Spurious Correlations in High Dimensional Regression: The Roles of Regularization, Simplicity Bias and Over-Parameterization

Neural Collapse Beyond the Unconstrained Features Model: Landscape, Dynamics, and Generalization in the Mean-Field Regime

Test-Time Training Provably Improves Transformers as In-context Learners