Approach to solve complex problems

My strategy for tackling complex problems: start simple. Though it might sound trivial, I’ve frequently noticed that many don't apply this approach.

In a recent discussion during my class, we explored the intricacies of debugging neural networks—taking the example of an object detector that outputs both the class and the bounding box coordinates. The approach I advocate is to start with simplicity: E.g. initially focus on the class output alone before tackling the regression of bounding box coordinates.