Many thanks in advance for taking time to read this question.
Recently, as a group of students, we reviewed a certain topic in Machine Learning. The supervisor asked us to prepare a survey paper on the topics that we have read. We have gathered the required content, reviewed individual papers and made a report. Now, we have touched around 50 pages, whereas the supervisor wants us to shorten the stuff. The journal which we are targeting, says that the survey papers/review papers should be less than 21 pages. The journal is from IEEE, and has two column format.
The content itself is heavy in terms of equations. For instance, If we write a review of a single paper, its going to take more than half of one page where we explain the underline methodology. That is, how they did the stuff. For remaining part of the page, we explain their experiments etc. For some papers, the methodology discussion takes bit more space, due to equations. It took us around 30-35 pages for reviews only. We also created a section for Background Knowledge, which covers around 8-10 pages. The total length goes to 50 pages, after we added analysis and discussion part.
Some of us have published papers already, however this particular field is a bit different for us. We understand to shorten the content, but we really dont know how to deal with this kind of papers having dozens of equations. We however have already picked the key ones, and left out the others. For instance:
My first question is related to Background Knowledge. Upto what extent one should describe a certain term. Are there any rules which should be considered while designing this section? like 20% of the whole paper?
My second question is particular to a single review (explanation of an article in my review paper) like the one I have shown above. For instance, in this particular paper, the author used an additional prior to handle feature sparsity. Now, I can put a single sentence to explain this, however, I believe that I put the equation there so that its clear, how they did it. In the same way, few more equations like likelihood definition, posterior form, and estimation equations might take some space. The whole thing swallows up. If there are like 20 such papers, obviously, it will end up with 35 pages, like it has happened. Should I leave out the equations? My question, is how should I setup the review format, what to include and what to exclude.
While reading the above, please note that I understand that in review papers, people are mostly interested in tables, comparison sheets and analysis stuff. However, we are still required to present enough review of papers.