Deep learning and Information Geometry

• Simplifying Momentum-based Positive-definite Submanifold Optimization with Applications to Deep Learning, ICML 2023, 2302.09738
• Tractable structured natural-gradient descent using local parameterizations, ICML 2021 2102.07405
  2107.10884
   
  
• Towards Modeling and Resolving Singular Parameter Spaces using Stratifolds, NeurIPS OPT 2021 workshop
   
  
• Sinkhorn AutoEncoders, UAI'19 1810.01118
   
  
• [IEEE TNNLS] 1806.00149 q-Neurons: Neuron Activations based on Stochastic Jackson's Derivative Operators, IEEE TNNLS 2020.
  CoLab
  
  
   
  
• [Neural Computing and Applications ] RG Anticipation-RNN: enforcing unary constraints in sequence generation, with application to interactive music generation, Neural Computing and Applications, 2018.
   
  
• [IEEE SSCI] arxiv GLSR-VAE: Geodesic latent space regularization for variational autoencoder architectures, IEEE Symposium Series on Computational Intelligence 2017
   
  
• JMLR proc DeepBach: a Steerable Model for Bach Chorales Generation, ICML 2017
  arxiv DeepBach: a Steerable Model for Bach Chorales Generation.
   
  
• JMLR proc Relative Fisher Information and Natural Gradient for Learning Large Modular Models, ICML 2017.
  arxiv Relative Natural Gradient for Learning Large Complex Models, 2016
  Web page
• Lightlike Neuromanifolds, Occam's Razor and Deep Learning
• q-Neurons: Neuron Activations Based on Stochastic Jackson's Derivative Operators
   
  
• arxiv Deep rank-based transposition-invariant distances on musical sequences, 2017