Below are some tips, mostly from other people, that I’ve compiled over my limited experience in writing research papers. I like Andrej Karpathy and Tom Cormen’s writing.
- Read some top papers in the conference/journal before you begin.
- The first sentence in a paragraph should tell me what the point of the paragraph is.
- Include math equations if relevant.
- You should make your code and data available, if possible.
- After you finish your first draft, go through the entire paper and delete any unncessary words. Also check all the prepositions to make sure they’re correct.
- It is good to have some sort of baseline approach for comparison, even if the point of your paper isn’t the methods. This came up a lot in reviewer comments for me when I didn’t include it.
- For methods papers, you should have a table comparing your results to other papers for a given benchmark task. You should also validate on multiple datasets.
- If you haven’t done an extensive literature review, reviewers will likely notice. If you miss a related paper that the reviewer happens to personally know of, your credibility immediately drops and you’ll likely get rejected.
- If your testing set is small, are you overfitting?
- If you’re presenting a new application of deep learning, do some error analysis. What is your model doing well and what isn’t it?
- What’s the training time and runtime of your model?
- Can you do a cute t-SNE visualization?
- Please, please use LaTeX. It is important that the look and feel of your paper is correct.
- Try a creative title. I should not get tired when I read it. Tom Cormen suggests titles like ‘Algol 68 with Fewer Tears’, ‘ViC*: Running out-of-memory instead of running out of memory”, ‘50 shades of grey codes’.
- Make sure your paper is exactly the page limit and not a single line less. Some tricks: delete spacing between figures and their captions, make the font in tables smaller.
- If your paper isn’t at the page limit, this could be an indicator that you should do more experiments.
- Include a pull figure on the first page to give the reader an idea of what the paper will be about.
- It’s often nice to have a table that summarizes the dataset(s) you are using. This is a must if you’re using a novel dataset.
- For vision applications, it’s always nice to have some LIME/CAM visualization to see what your model is looking at.
- For multi-class precision/recall/f1 tables, try a plot like in Figure 3B here.
- Include ROC and AUC if applicable, and show the point on the curve that the model is performing at.
- You can display a confusion matrix as a heatmap.
- For a multivariate function z = f(x, y), try using a heatmap.
Some phrases that I like:
- “We propose/present a deep learning model that xyz.” -Andrej Karpathy
- “Of note, we are the first to xyz.” -Saeed Hassanpour
- “In closing, we xyz.”
- “Board-certified” radiologist/pathologist sounds good for asserting dominance.
- Transition words I like: in addition, furthermore, moreover.
Easiest reasons for a reviewer to reject you:
- No literature review
- No methodological innovation
Need more inspiration?: