A study of using GANs to build a simulation environment from screen information alone. There are 3 module. Action conditioned LSTM incorporating, External memory read/write mechanism to record the entire map like NTM, decoder for predicting the next frame from those.
TL;DR
A study of using GANs to build a simulation environment from screen information alone. There are 3 module. Action conditioned LSTM incorporating, External memory read/write mechanism to record the entire map like NTM, decoder for predicting the next frame from those.
Why it matters:
Paper URL
https://arxiv.org/abs/2005.12126
Submission Dates(yyyy/mm/dd)
Authors and institutions
Methods
Results
Comments