LinGCN: Structural Linearized Graph Convolutional Network for Homomorphically Encrypted Inference

Keyphrases

• Homomorphic Encryption 100%
• Graph Convolutional Network 100%
• Encrypted Inference 100%
• Latency 75%
• Client Data 50%
• Distillation 25%
• Network Model 25%
• Privacy-preserving Machine Learning 25%
• Financial System 25%
• Large Animal Model 25%
• Activation Function 25%
• Two-level 25%
• Accuracy Improvement 25%
• Computation Overhead 25%
• Privacy Concerns 25%
• Human Performance 25%
• Personal Health Systems 25%
• Teacher Model 25%
• Node Level 25%
• Inference Accuracy 25%
• Model Weight 25%
• Skeleton Joints 25%
• Linear Position 25%
• Model Size 25%
• Compact Node 25%
• Indicator Function 25%
• Replacement Policy 25%
• Location Selection 25%
• Network Inference 25%
• Adversarial Attack 25%
• GitHub 25%
• Security Concerns 25%
• Rectified Linear Unit (ReLU) 25%
• Optimization Goal 25%
• Trainable Activation Functions 25%
• Operator Fusion 25%
• Linearization Algorithm 25%

Computer Science

• Homomorphic Encryption 100%
• Graph Convolutional Network 100%
• Data-Client 66%
• Activation Function 66%
• Privacy Concern 33%
• Security Concern 33%
• Network Inference 33%
• Replacement Policy 33%
• Adversarial Machine Learning 33%
• Human Performance 33%
• Indicator Function 33%
• Privacy Preserving Machine Learning 33%
• homomorphic-encryption 33%
• Encryption Solution 33%