On the expressive power of deep learning: A tensor analysis

Mathematics

• Expressive Power 100%
• Learning 78%
• Tensor 69%
• Pooling 32%
• Locality 29%
• Architecture 27%
• Sharing 25%
• Hierarchical Networks 19%
• Network Architecture 18%
• Arithmetic Circuits 16%
• Decompose 15%
• Measure Theory 14%
• Beliefs 13%
• Matrix Algebra 13%
• Community 12%
• Justification 12%
• Text 11%
• Evidence 10%
• Factorization 10%
• Transform 8%
• Equivalence 8%
• Performance 7%
• Polynomial 6%

Engineering & Materials Science

• Tensors 85%
• Deep learning 76%
• Decomposition 41%
• Factorization 28%
• Network architecture 26%
• Polynomials 22%
• Set theory 18%
• Networks (circuits) 15%