Multi-Step Span Loss Prediction in Optical Networks Using Multi-Head Attention Transformers

Keyphrases

• Transformer 100%
• Optical Networks 100%
• Window Size 100%
• Multi-head Attention 100%
• Loss Prediction 100%
• Recurrent Neural Network 80%
• Linear Regression 80%
• Maximum Absolute Errors 60%
• Absolute Maximum 60%
• Single-step Model 60%
• Root Mean Square Error 40%
• Forecast Horizon 40%
• Naive Model 40%
• Naive Approach 40%
• Multi-step Model 40%
• Model Performance 20%
• Prediction Accuracy 20%
• Deep Learning Model 20%
• Performance Monitoring 20%
• Optimal Performance 20%
• Prediction Task 20%
• Proactive Monitoring 20%
• Self-attention Mechanism 20%
• Multi-step Prediction 20%
• Linear-in-dB 20%
• Error Improvement 20%
• Model Average 20%
• Transformer-XL 20%
• Long Model 20%

Engineering

• Fiber Optic Networks 100%
• Multi-Head Attention 100%
• Step Model 100%
• Recurrent Neural Network 80%
• Maximum Absolute Error 60%
• Root Mean Square Error 40%
• Deep Learning Method 20%
• Field Data 20%
• Accurate Prediction 20%
• Optimal Performance 20%
• Comparable Result 20%
• Self-Attention Mechanism 20%