Synthetic data as a revolutionary solution for rare data
Rare data occurs in the real world because of infrequent events or phenomena. In other words, these events occur but with low frequency. We can broadly classify these events into these categories:
- Natural catastrophes: This category includes events such as floods, asteroid impacts, earthquakes, and tsunamis
- Anthropogenic: This category includes events such as industrial accidents, financial crises, and violent conflicts
These events create many major changes in the environment, which may cause state-of-the-art ML models to fail. For example, a face recognition system may not work well in the case of the evacuation of a building because as the building becomes more crowded, movement patterns may change. While these events are rare, their impacts on societies are tremendous. ML models that function inappropriately may greatly increase the number of deaths and injuries.
For ML models to be robust and accurate...
