Philipp Petersen
Scholar

Philipp Petersen

Google Scholar ID: Huw7cHIAAAAJ
University of Vienna
Applied Harmonic AnalysisDifferential equationsNeural network approximation
Citations & Impact
All-time
Citations
3,620
 
H-index
21
 
i10-index
27
 
Publications
20
 
Co-authors
11
list available
Contact
Resume (English only)
Academic Achievements
  • Co-authored preprint 'Exponential ReLU Neural Network Approximation Rates for Point and Edge Singularities' with Carlo Marcati et al., demonstrating exponential approximation rates of neural networks on general bounded domains
  • Co-authored preprint 'Neural network approximation and estimation of classifiers with classification boundary in a Barron class' with Andrei Caragea and Felix Voigtlaender, showing dimension-independent approximation rates for functions with Barron-type singularities
  • Collaborated with Gitta Kutyniok, Mones Raslan, and Reinhold Schneider on a preprint showing that neural network size for approximating parametric maps depends on reduced basis size rather than discretization dimension, overcoming the curse of dimensionality
  • Co-authored preprint 'Equivalence of approximation by convolutional neural networks and fully-connected networks' with Felix Voigtlaender, establishing theoretical equivalence in approximation capabilities between CNNs and fully-connected networks
  • Co-wrote a humorous paper 'The Oracle of DLPhi' highlighting the need for skepticism in mathematical claims about deep learning