Best Practices for Choosing the Number of Iterations

[Golbstein]

Is there a general guideline for determining the optimal number of iterations?

For example, with a dataset of 500 images, running 30k iterations results in each image being "seen" approximately 60 times. Is this considered ideal? I have a dataset with 8,000 images, and 30k iterations seem insufficient.

Should I rely on trial and error until the results are satisfactory, or are there any comprehensive studies or best practices on this topic?

I'd appreciate any insights or experiences others have had!

[leblond14u]

From what I seen and experienced so far it seems to be mostly trial and error.

As there's additional parameters to factor in this choice, like the scale of the dataset etc, that can greatly influence the number of iterations you would want to chose.

Best,

Hugo