PDF Publication Title:
Text from PDF Page: 046
Fig 5. Combination of two-layers CNN and two-layers RNN model. Architecture of the best CNN-LSTM network applied to QoE prediction The architecture presented in Fig 5 corresponds to the best one obtained in [4] for QoE prediction. Figs 6-8 present other possible configurations with different number of CNN and LSTM layers. Fig 8 shows the architecture that provides best results for the research performed on NTC prediction [2]. We can see that the number of layers has an impact on the results obtained and that increasing this number does not always improve them. In addition to the layers shown in Figs 5-8, other layers have been considered for this architecture: Max Pooling, Batch Normalization and Drop-out layers [2][4]. Max pooling provides a further dimensionality reduction in the generated tensors apart from additional properties of invariance. Batch normalization improves the stability of the learning process and, sometimes, makes it faster. Regularization using drop-out is a very effective way to avoid over-fitting for neural network architectures. These additional layers were tested in the experiments conducted for the thesis, but were not included in the final models, with the exception of the drop-out layer that was used to improve generalization (avoid overfitting). Fig. 6. Simple one-layer RNN model Doctoral Thesis: Novel applications of Machine Learning to NTAP - 44PDF Image | Novel applications of Machine Learning to Network Traffic Analysis
PDF Search Title:
Novel applications of Machine Learning to Network Traffic AnalysisOriginal File Name Searched:
456453_1175348.pdfDIY PDF Search: Google It | Yahoo | Bing
Cruise Ship Reviews | Luxury Resort | Jet | Yacht | and Travel Tech More Info
Cruising Review Topics and Articles More Info
Software based on Filemaker for the travel industry More Info
The Burgenstock Resort: Reviews on CruisingReview website... More Info
Resort Reviews: World Class resorts... More Info
The Riffelalp Resort: Reviews on CruisingReview website... More Info
| CONTACT TEL: 608-238-6001 Email: greg@cruisingreview.com | RSS | AMP |