create new data points that are similar to the existing data but have been modified in some way that add diversity. Add these in manner to increase robust ness of model. Make your data representative of real world. Ensure your model will be able to handle situations it has not seen yet.
Use Machine learning and a generative model approaches to create new data points that are made up of combinations of existing data points. It solve data sparsity problem. Using knobs ensure that new data points are added in compliant manner and bis is not introduced.
Combining existing data points in new ways. This can be done by using a generative model to create new data points that are made up of combinations of existing data points. Example Combine photos taken for different car models in day, night condition can be used to produce bigger dataset with all car models in all conditions. Similarly one can use text conversation for multiple product and combine better dataset.
Modifying existing data points. This can be done by using a generative model to create new data points that are similar to the existing data but have been modified in some way. Example - text written can be enhanced with short sentences, bigger sentences, sentences are like questions.
Technology : Optimal Transport, Generative AI models like GAN, Active Learning
Discovering abstractions is crucial to reduce the amount of experience or thinking time.
Using knobs you can manage a trade off between compression of states and representation of good behavior.