Research Report Content

[TUe-Ray]

I have a few questions regarding the final report.

Since I tried a few different methods and architectures, would it be helpful (or expected) to include those exploration steps and findings in the report? Or should we just focus on the final model and the best optimization results?

Also, some of the experiments or optimization techniques were tested on earlier model versions and then carried over to the newer architecture without re-running. Should I rerun those on the final model to include the results, or is it okay to just explain that in the write-up?

One last thing—how should I share the GitHub repo with you?

[TimJaspers0801]

In general, I’d recommend including most of your results in the report. Just try to organize them logically—group similar comparisons together. For example, you could have one table comparing different model architectures, and another focusing on optimization techniques. Eventually, it would make sense to train a final model using the best-performing architecture and optimization method.

As for the GitHub repository, you can simply make it public and include the link in your report so we can take a look.

Good luck!