GDPP: Learning Diverse Generations using Determinantal Point Processes

Keyphrases

• Determinantal Point Processes 100%
• Diverse Generation 100%
• Generative Models 60%
• Mode Collapse 60%
• Synthetic Data 40%
• Probability Distribution 20%
• Arts-based Methods 20%
• Data Distribution 20%
• Input Noise 20%
• Convergence Time 20%
• Image Data 20%
• High-dimensional Probability 20%
• Data Efficiency 20%
• Data Image 20%
• Training Scheme 20%
• Quality Time 20%
• High-quality Sample 20%
• Training Paradigm 20%
• Training Parameters 20%
• Natural Images 20%
• MNIST 20%
• Variational Autoencoder 20%
• Point Process Approach 20%
• Adversarial Training 20%
• CIFAR-10 20%
• Multimodal Output 20%
• Probabilistic Measures 20%
• Few-mode 20%
• Complex Training 20%
• Parameters Training 20%
• Unsupervised Penalty 20%

Computer Science

• Generative Model 100%
• Synthetic Data 66%
• Autoencoder 33%
• Negative Correlation 33%
• Essential Characteristic 33%
• Adversarial Machine Learning 33%
• Data Distribution 33%
• Process Approach 33%

Engineering

• Point Process 100%
• Generative Model 60%
• Real Data 40%
• State-of-the-Art Method 20%
• Convergence Time 20%
• Natural Image 20%
• Loss Penalty 20%
• High-Dimensional Probability 20%
• Input Noise 20%
• Process Approach 20%
• Autoencoder 20%