Generalized Depthwise-Separable Convolutions for Adversarially Robust and Efficient Neural Networks

Keyphrases

• Depthwise Separable Convolution 100%
• Robust Neural Networks 100%
• Efficient Neural Network 100%
• Depthwise Separable 100%
• Model Compression 50%
• Adversarial Perturbation 50%
• 2D Convolution 50%
• Frames per Second 25%
• Compression Techniques 25%
• Exact Algorithm 25%
• Training Model 25%
• Additional Training 25%
• Large Problem Size 25%
• Complex Constraints 25%
• Convolutional Neural Network 25%
• Storage Cost 25%
• Computational Storage 25%
• Error Constraint 25%
• ImageNet Dataset 25%
• CIFAR-10 25%
• Adversarial Training 25%
• Pre-trained Networks 25%
• Combining Models 25%

Computer Science

• Neural Network 100%
• Depthwise Separable Convolution 100%
• Model Compression 66%
• Computer Hardware 66%
• Approximation (Algorithm) 33%
• Compression Technique 33%
• Large Problem Size 33%
• Exact Algorithm 33%
• Adversarial Machine Learning 33%
• Trained Network 33%
• Convolutional Neural Network 33%
• Pre-Trained Model 33%

Mathematics

• Neural Network 100%
• Convolution 100%
• Real Life 40%
• Optimality 20%
• Imagenet Dataset 20%
• Convolutional Neural Network 20%