Synthetic data: data balancing and data augmentation

Artificial intelligence (AI) based software from Syntho will be used to reduce ‘algorithmic biases’ and discrimination. We do this by supplementing shortages of training data, whereby certain groups of people are underrepresented. We can correct colored data from the past, whereby certain groups of people are negatively underrepresented or overexposed, through the generation of ‘fair’ synthetic data. In the figure above, we see an example of an input dataset with an improper balance between genders at the left side of the illustration. After training the Syntho Engine on this data and correcting for imbalances, a new synthetic dataset is generated that balances males and females, as illustrated at the right side of the image. In conclusion, this allows for smart data balancing by generating intelligent synthetic data.