It's about time. Modelling human visual inference with deep recurrent neural networks.

Tim Kietzmann, Donders Institute for Brain, Cognition and Behaviour

[iuhh]

Fascinating result on how the latent representation of picture and words describing the picture are so close. And from this I am thinking the memory retrieval in the brain is like feeding this latent representation into a decoder of specific modal, perhaps then ran through a GAN like network to ensure correct reconstruction - like a face in your head that you can't quite put your finger on it and keep trying to "remember" what it looked like. Very keen to learn more on this topic, thanks for sharing.