This is done by B.SatyaKishore under the guidance of Dr.Tapan K Gandhi at IIT, Delhi as a part of M. Tech Thesis. We here created a training process which won’t seem random to autistic people and thus helps them get trained. We tested these claims on Okay? game available in play and app store.
The main idea is we divided the decision making process into two parts: - Action Choosing Process [Markov Decision Process] - Knowledge Encoding
We then approximated the errors in Knowledge Encoding using Gaussian Distribution and tried to find the Knowledge encoing of autistic person by argmax over encodings based on the actions taken.Project is available at GITHUB LINK but is private.