FedMM: A Communication Efficient Solver for Federated Adversarial Domain Adaptation

Keyphrases

• Minimax 100%
• Stochastic Gradient Descent 100%
• Communication Efficiency 100%
• Unlabeled Data 100%
• Efficient Solvers 100%
• Adversarial Domain Adaptation 100%
• Divergence 50%
• Accuracy Improvement 50%
• Heterogeneous Data 50%
• Multiple Steps 50%
• Stationary Point 50%
• Benchmark Dataset 50%
• Training Model 50%
• Adversary 50%
• Averaging Method 50%
• Domain Adaptation 50%
• Dual Variable 50%
• Satisfactory Solution 50%
• Target Domain 50%
• Ascent Algorithm 50%
• Training Tasks 50%
• Local Update 50%
• Source Domain 50%
• Communication Round 50%
• Minimax Optimization 50%
• Training from Scratch 50%

Computer Science

• Domain Adaptation 100%
• Unlabeled Data 66%
• Gradient Descent 66%
• Stationary Point 33%
• Heterogeneous Data 33%
• Satisfactory Solution 33%
• Pre-Trained Model 33%