PDF Publication Title:
Text from PDF Page: 172
Figure 4. Option B. VGM model with Gaussian and Bernoulli distributions plus RMSE for continuous features. 3.3.3. Option C. Model based on Conditional VAE with Gaussian and Bernoulli distributions The last option is presented in Figure 5. This option is similar to option A, but with an important difference. Instead of using the labels as the input for the encoder we use it as an additional input to the decoder. We leave the features vector (𝑿) as input to the encoder. The architecture is similar to a normal VAE except for the inclusion of the label vector as a supplementary input to the decoder. In order to get the label vector inside the decoder network we just concatenate it with the values of the first layer of the decoder block (Figure 5). The one-hot encoded label vector has a length of 5 that is too short to have a significative impact in the output, which is why we replicate the label vector 40 times before we concatenate it with the first decoder layer. Then, what is represented in Figure 5 as the vector L is really a vector of length 200 (the label vector replicated 40 times). The number of times we replicate the label vector can be considered as a hyperparameter of the model: this number modifies the results but not in a very significant way; it can be considered as a part of the fine tuning of the model. In Figure 5, in squared boxes are the elements of the loss function to be minimized for this option. This option has produced the best results. Doctoral Thesis: Novel applications of Machine Learning to NTAP - 170PDF 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 |