A QoS Classifier Based on Machine Learning for Next-Generation Optical Communication

Round 1

Reviewer 1 Report

the title may mislead the reader to think that the paper will present hardware related design and development I suggest a change e.g. An QoS Classifier Based on Machine Learning for Next Generation Optical Communication

also to remove from keywords High-Performance Embedded Systems from the same reasons

or provide supplementary information related to proposed hardware that will implement the ml 

 

Author Response

Reviewer #1

----------------------

Dear Reviewer, Thank you very much for your positive response and beneficial feedback. The following are the reviewer comments and the corresponding response.

1. 
   The title may mislead the reader to think that the paper will present hardware related design and development I suggest a change e.g. An QoS Classifier Based on Machine Learning for Next Generation Optical Communication also to remove from keywords High-Performance Embedded Systems from the same reasons or provide supplementary information related to proposed hardware that will implement the ml.

Dear Reviewer, many thanks for your comment.

You are right. The paper name is changed to 'An QoS Classifier Based on Machine Learning for Next Generation Optical Communication'. Moreover, the keywords related to hardware are removed and replaced by 'Hypothesis Testing'.

Author Response File: Author Response.pdf

Reviewer 2 Report

1. The introduction should adequately define the proposed solution's purpose within 6G networks. So, the state where or in what place in the 6G network you are implementing the proposed solution. What effects can it have on other entities in 6G networks?

2. The Related works section is missing, which is critical to see if there is related research.

3. A more detailed description of the ML experiment conducted is missing.

4. A more detailed description of the ML experiment conducted is missing.
The story should be introduced much deeper into the context of 6G networks. There are defined KPIs for 6G networks. This solution must be considered and how the proposed solution affects the achievement of KPI goals.

5. The authors should improve the conclusion and the reference list.

Author Response

Reviewer #2

----------------------

Dear Reviewer, Thank you very much for your positive response and beneficial feedback. The following are the reviewer comments and the corresponding response.

1. 
   The introduction should adequately define the proposed solution's purpose within 6G networks. So, the state where or in what place in the 6G network you are implementing the proposed solution. What effects can it have on other entities in 6G networks?

Dear Reviewer, many thanks for your comment.

One of the important possible applications are the generation passive optical networks (PONs). Besides, the proposed solution is addressed in QoS which requires achieving certain KPIs requirements.

-------------------------------------------------------------------------------------------------------------

2. The Related works section is missing, which is critical to see if there is related research.

Dear Reviewer, many thanks for your comment.

The related works are shown in the introduction in references [5-9]. The work done is novel work as it deals with classification in QoS.

-------------------------------------------------------------------------------------------------------------

3. A more detailed description of the ML experiment conducted is missing.

Dear Reviewer, many thanks for your comment.

The section of the ML procedures is modified in order to show the problem formulation and conduction.

The ML problem formulation is to classify between different data rates at different detection techniques (DD and SPD) to achieve a certain key performance indicator (KPI) need. The Q, SNR and BER are used as ML predictors. The DT and RF are both used as classifiers. The classification process is formulated in order to determine four different classes with different data rates {DD: GROUPS A/D, DD: GROUPS B/C, SPD: GROUPS A/D, SPD: GROUPS B/C}. Hence, the classifiers are used as supervised learning where the training data has an output included while training the model. The total dataset rows and columns are 1800 and 4, respectively. The four columns are the Q, SNR and BER while the fourth column is the label outcome. The number of training data points is 1200 (300 data point for each class of the four classes). Then, the number of test data points is 600 (150 data point for each class of the four classes).

-------------------------------------------------------------------------------------------------------------

4. The story should be introduced much deeper into the context of 6G networks. There are defined KPIs for 6G networks. This solution must be considered and how the proposed solution affects the achievement of KPI goals.

Dear Reviewer, many thanks for your comment.

The KPIs are introduced in ML procedures section.

The ML problem formulation is to classify between different data rates at different detection techniques (DD and SPD) to achieve a certain key performance indicator (KPI) need. The KPIs needs are the following spectrum, we use C-band we can use it in next generation passive optical network ( PON),  data rates (1.25 Gbps for Audio and 2.5 Gbps for video). The Q, SNR and BER are used as ML predictors. The DT and RF are both used as classifiers. The classification process is formulated in order to determine four different classes with different data rates {DD: GROUPS A/D, DD: GROUPS B/C, SPD: GROUPS A/D, SPD: GROUPS B/C}. Hence, the classifiers are used as supervised learning where the training data has an output included while training the model. The total dataset rows and columns are 1800 and 4, respectively. The four columns are the Q, SNR and BER while the fourth column is the label outcome. The number of training data points is 1200 (300 data point for each class of the four classes). Then, the number of test data points is 600 (150 data point for each class of the four classes).  

-------------------------------------------------------------------------------------------------------------

5. The authors should improve the conclusion and the reference list.

Dear Reviewer, many thanks for your comment.

The conclusion is modified. The reference list is updated by modifying the references: [1, 2, 8, 13, 15, 16, 17, 18].

 

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Thank you for accepting the remarks and correcting the article.