Here, we present a novel approach to solve the problem of reconstructing perceived stimuli from brain responses by combining probabilistic inference with deep learning. Our approach first inverts the linear transformation from latent features to brain responses with maximum a posteriori estimation and then inverts the nonlinear transformation from perceived stimuli to latent features with adversarial training of convolutional neural networks. We test our approach with a functional magnetic resonance imaging experiment and show that it can generate state-of-the-art reconstructions of perceived faces from brain activations.

The authors propose a brain decoding model tailored to face reconstruction from BOLD fMRI measurements of perceived faces. There are some promising aspects to this contribution, but overall in its current state there are also a number of concerning issues. Positive points: - a GAN decoder was trained on face embeddings coming from a triplet loss or identity-predicting face embedding space to output the original images. Modulo my inability to follow the deluge of GAN papers closely, this is a novel contribution in that it is the application of the existant imagenet reconstruction GAN to faces. This itself may be on the level of a workshop contribution.

Here, we present a novel approach to solve the problem of reconstructing perceived stimuli from brain responses by combining probabilistic inference with deep learning. Our approach first inverts the linear transformation from latent features to brain responses with maximum a posteriori estimation and then inverts the nonlinear transformation from perceived stimuli to latent features with adversarial training of convolutional neural networks. We test our approach with a functional magnetic resonance imaging experiment and show that it can generate state-of-the-art reconstructions of perceived faces from brain activations. Papers published at the Neural Information Processing Systems Conference.