Iterative refinement in the continuous space for non-autoregressive neural machine translation

Keyphrases

• Autoregressive Model 33%
• BLEU Score 33%
• Computation Overhead 33%
• Computationally Efficient 33%
• Continuous Latent Variable 33%
• Continuous Space 100%
• Efficient Inference 33%
• Ende 66%
• Gradient-based Optimization 33%
• Hybrid Space 33%
• Inference Methods 100%
• Inference Network 33%
• Inference Time 33%
• Iterative Refinement 100%
• IWSLT 33%
• Latent Space 33%
• Latent Variable Models 33%
• Latent Variables 33%
• Low Dimensionality 33%
• Machine Translation 66%
• Marginal Probability 66%
• Non-autoregressive 66%
• Non-autoregressive Neural Machine Translation 100%
• Probability Score 33%
• Token 33%
• Translation Quality 33%

Computer Science

• Continuous Latent Variable 33%
• Continuous Space 100%
• Continuous Variable 33%
• Discrete Variable 33%
• Iterative Refinement 100%
• Latent Variable Model 33%
• Machine Translation 100%
• Marginal Probability 66%
• Network Inference 33%
• Refinement Step 100%
• translation quality 33%

Mathematics

• Approximates 33%
• Autoregressive Model 33%
• Continuous Variable 33%
• Discrete Variable 33%
• Inference Procedure 100%
• Iterative Refinement 100%
• Latent Variable Model 33%
• Marginal Probability 66%
• Marginals 33%
• Probability Theory 33%
• Refines 66%