Learning predictable and robust neural representations by straightening image sequences | Read Paper on Bytez