I would like to ask to the community about a very common problem that is not being discussed in the topics of the course.
Which is the best aproach in case your dataset is imbalanced? To give a simple example, suppose an image binary classification dataset of dogs and cats in which you have a lot more images of dogs compared than cats.
Could we train with this data set as it is?, or better we should reduce the number of dogs to equal the number of cats.
There is another approach we must try?
Thanks in advance