Deep Learning (DL) attempts to track down the unobtrusive, covered-up connections (assuming any!) between different input features (which could add up to a large number). These connections are non-self-evident. While this inquiry is certifiably not a thorough pursuit, DL needs to investigate bunches of potential outcomes.
The issue is when DL investigates a lot of potential outcomes, the chances of DL finding a fake relationship that happens by chance increments. By expanding the size of the preparation set, you decrease the chances of DL finding fake connections just as increment the chances of tracking down the unobtrusive connections that are truly there.
Interested to learn Machine learning in detail? Come and Join the Machine learning course.