Export Reviews, Discussions, Author Feedback and Meta-Reviews

First provide a summary of the paper, and then address the following criteria: Quality, clarity, originality and significance. Authors propose an attentional neural network that integrates top-down cognitive bias and bottom-up feature extraction. When there is noise in the stimulus, it can resolve the confusion by setting the appropriate top-down bias. They apply their method for digit recognition on noisy images (random background and random image) as well as two overlaid digits and report high accuracies in par or better than state of the art. I found this result interesting but I am not quite sure of the novelty of the approach specially considering a volume of research in recurrent networks, Hopfield nets, predictive coding, etc.