Why skewed data increases error after few iterations



I was not able to understand answer of Q40 from https://www.analyticsvidhya.com/blog/2017/01/must-know-questions-deep-learning/. Why would skewed data increases the error steeply after few iterations



skewed data is something when you have one class of data in large amount and other class in small quantities.
So, the machine learning algorithm is not able to build a proper rule for classes with lesser data.

In this case, your model might work in few iteration of train and test. However, when you iterate over and validate your model with different test data, your accuracy hampers a lot. because the new itertion of test data might not follow the same distribution of your skewed training data