Course talk:CPSC522/Network Agent

Comments

The page is focused, comprehensive, and well-written; it was easy to read along despite the amount of information in the article. The problem statement and the approach of the project are clear. Since the work is an ongoing research project, it is understandable that the results are not as impressive yet, but seems quite promising. Personally I find the problem of network routing very interesting, so this article was enjoyable.

Minor comments

• In Experiment/Methodology section, I believe there is a typo:

"In our experiment we continuously send UDP traffic from H1 to H3 and from H3 to H4."

should be

"In our experiment we continuously send UDP traffic from H1 to H3 and from H2 to H4."

• The text in Experiment/Results section is hard to read because of the large graph. Maybe put the text below the graph and let the graph take up the whole horizontal space?

• The graphs in Experiment/Results sections are not difficult to interpret (as the axes are clearly labeled), but it wouldn't hurt to briefly describe at the beginning of the section what each of the three graphs (per agent) display, as the text on the graph are very small. For example, first graph shows reward over time - higher the better, second graph shows queuing delay - lower the better, etc.

Scheme

• The topic is relevant for the course. 5
• The writing is clear and the English is good. 5
• The page is written at an appropriate level for CPSC 522 students (where the students have diverse backgrounds). 5
• The formalism (definitions, mathematics) was well chosen to make the page easier to understand. 5
• The abstract is a concise and clear summary. 5
• There were appropriate (original) examples that helped make the topic clear. 4
• There was appropriate use of (pseudo-) code. -
• It had a good coverage of representations, semantics, inference and learning (as appropriate for the topic). 4
• It is correct. 5
• It was neither too short nor too long for the topic. 4
• It was an appropriate unit for a page (it shouldn't be split into different topics or merged with another page). 5
• It links to appropriate other pages in the wiki. 4
• The references and links to external pages are well chosen. 4
• I would recommend this page to someone who wanted to find out about the topic. 5
• This page should be highlighted as an exemplary page for others to emulate. 5
• If I was grading it out of 20, I would give it: 18

Thanks, great suggestions!

The topic is interesting and a good fit for this course. I would like to see more explanation on your motivation for choosing the 4 different agents that you tried.

• The topic is relevant for the course. 5
• The writing is clear and the English is good. 5
• The page is written at an appropriate level for CPSC 522 students (where the students have diverse backgrounds). 5
• The formalism (definitions, mathematics) was well chosen to make the page easier to understand. 5
• The abstract is a concise and clear summary. 5
• There were appropriate (original) examples that helped make the topic clear. 5
• There was appropriate use of (pseudo-) code. N/A
• It had a good coverage of representations, semantics, inference and learning (as appropriate for the topic). 4
• It is correct. 5
• It was neither too short nor too long for the topic. 4
• It was an appropriate unit for a page (it shouldn't be split into different topics or merged with another page). 5
• It links to appropriate other pages in the wiki. 5
• The references and links to external pages are well chosen. 5
• I would recommend this page to someone who wanted to find out about the topic. 4
• This page should be highlighted as an exemplary page for others to emulate. 4

I would give this page a 19/20

Thanks! To be honest, the algorithms we chose are largely random. We were mostly were trying out stuff. But will try to provide explanations.

Like the other review said this article is well written, balanced and focused. I would like to see an explanation of why reinforcement learning is a good method for routing? In my understanding, reinforcement learning is useful when the environment responds with a score/reaction, but the agent has no knowledge of how the environment has come to that reaction. In the case of network routing you do have access to total and individual bandwidth, etc...

Thanks for the feedback, tried to address it.