Parent: Data in ML

Source: google-ml-course

Sampling data in ML

Three basic assumptions

1. Data points are drawn independently and identically (i.i.d.), and at random, from the data distribution, i.e. the data points don’t influence each other
2. The data distribution doesn’t change over time (stationary)
3. The data is pulled from the same distribution, for both training and validation sets

Situations where these assumptions may be violated:

• change in user perception, therefore resulting in different labelling of a dataset
• change in population which result in new demographics or target market