Problem with Tip 4 -- Throw out noisy data

My concern about throwing out noisy data is that it makes the model less robust. In the real world, data is going to have noise – for example, images won’t always be perfectly focused.