Products of Many Large Random Matrices and Gradients in Deep Neural Networks

Mathematics

• Random Matrices 100%
• Neural Networks 94%
• Activation 78%
• Gradient 76%
• Tend 49%
• Finite Temperature 49%
• Products of Random Matrices 48%
• Moment 45%
• Infinity 45%
• Kolmogorov-Smirnov 43%
• Norm 40%
• Scaling Limit 38%
• Explicit Bounds 37%
• Architecture 36%
• Logarithm 32%
• Fluctuations 31%
• Error Bounds 29%
• Limiting 26%
• Term 17%
• Theorem 13%

Physics & Astronomy

• norms 73%
• infinity 66%
• gradients 65%
• products 64%
• matrices 54%
• activation 53%
• moments 48%
• logarithms 36%
• entry 35%
• theorems 27%
• scaling 23%
• temperature 9%