Learning to Balance Local Losses via Meta-Learning

Keyphrases

• Benchmark Dataset 25%
• CIFAR-10 25%
• CIFAR-100 25%
• Competitive Performance 25%
• Complex Deep Neural Networks 25%
• Deep Neural Network 50%
• Dropout 25%
• Dynamic Loss Function 25%
• Effective Training 25%
• Error Signal 25%
• Global Loss 25%
• Image Classification 25%
• Local Error 25%
• Local Loss 100%
• Loss Function 100%
• Meta-learning 100%
• Overfitting 25%
• Popular 25%
• Skip Connection 25%
• State-of-the-art Techniques 25%
• Training Deep Neural Networks 25%
• Training Dynamics 25%

Computer Science

• Deep Neural Network 100%
• Image Classification 25%
• Meta-Learning 100%