Adaptive learning rate for increasing learning speed in backpropagation networks

Keyphrases

• Back Propagation Network 100%
• Learning Rate 100%
• Learning Speed 100%
• Backpropagation 66%
• Back Propagation Algorithm 44%
• Training Set 22%
• Higher-order Derivatives 22%
• Quasi-optimal 22%
• Learning Parameters 22%
• Function Approximation 11%
• Computationally Expensive 11%
• Error Surface 11%
• Real-world Problems 11%
• Pattern Classification 11%
• Parameter Values 11%
• Weight Change 11%
• Computational Complexity 11%
• Elimination Method 11%
• Partial Derivatives 11%
• Multilayer Neural Network 11%
• Rate-based 11%
• Function Pattern 11%
• Mean Square Error 11%
• Flat Spot 11%
• Gradient Information 11%
• Training Epochs 11%
• Guesswork 11%
• Overall Error 11%
• Derivative Approximation 11%
• Modified Algorithm 11%
• Optimal Learning Rate 11%
• Reasonable Value 11%
• Supervised Learning Algorithms 11%
• Gradient Descent 11%
• Simple Heuristics 11%
• Second-order Derivative 11%
• Adaptive Modification 11%
• Local Gradient 11%
• Weighted Spaces 11%
• Error Metrics 11%

Computer Science

• Learning Rate 100%
• Learning Algorithm 30%
• Learning Parameter 20%
• Backpropagation Algorithm 20%
• Speed-up 10%
• Approximation (Algorithm) 10%
• Function Approximation 10%
• Pattern Classification 10%
• Supervised Learning 10%
• Input/Output 10%
• Computational Complexity 10%
• Real-World Problem 10%
• Gradient Descent 10%
• Modified Version 10%
• Elimination Method 10%
• Parameter Value 10%
• Layer Neural Network 10%
• Output Mapping 10%