Learning Sparse Boolean Functions: Neural Networks Need a Hierarchical Degree Chain
Explore neural networks' hierarchical degree chain for learning sparse Boolean functions. Gain insights from EPFL expert E. Abbé.
Benefits of Deep and Narrow Neural Networks for Approximation and Memorization
Explore groundbreaking research on neural networks' architecture, revealing why depth trumps width in function approximation and how deep, narrow networks can efficiently memorize data using minimal parameters.
Networks that Adapt to Intrinsic Dimensionality Beyond the Domain
Explore neural networks' ability to adapt to intrinsic dimensionality, focusing on ReLU networks approximating functions with dimensionality-reducing feature maps. Gain insights into manifold learning and data analysis.
The Perceptron as a Roadmap - From Neuron Structure to Artificial Neural Networks
Explore the evolution of neural networks from biological neurons to machine learning, using the perceptron as a guide. Uncover connections to associative memories, function approximation, and surprising mathematical links.
Toward Understanding the Extrapolation of Nonlinear Models to Unseen Domains
Explore advanced concepts in domain adaptation and model extrapolation, focusing on nonlinear models' behavior in unseen domains through theoretical and practical analysis.