ACLI: A CNN Pruning Framework Leveraging Adjacent Convolutional Layer Interdependence and γ-Weakly Submodularity

Keyphrases

• Pruning 100%
• Convolutional Neural Network 100%
• Convolutional Layer 100%
• Neural Network Pruning 100%
• Interdependence 100%
• Submodularity 100%
• Pruning Strategy 66%
• Importance Function 66%
• Order of Magnitude 33%
• High Efficiency 33%
• Network Parameters 33%
• Computational Cost 33%
• Easy-to-implement 33%
• Resource Efficiency 33%
• Low Quality Data 33%
• Absolute Error 33%
• L2 Norm 33%
• Efficiency Metrics 33%
• Source-free 33%
• Weak Submodularity 33%
• Submodular Function 33%
• Benchmark Network 33%
• Oblivious Algorithms 33%

Computer Science

• Convolutional Neural Network 100%
• Convolutional Layer 100%
• Interdependence 100%
• Network Parameter 50%
• Computational Cost 50%
• Theoretical Framework 50%
• Oblivious Algorithm 50%