Peng Liu

This paper introduces EmoFB, a deep neural network model that simulates emotional perception by integrating top-down feedback signals, showing improved recognition and alignment with human brain responses.

This paper introduces a biologically inspired deep neural network model for visual emotion processing and shows it accurately predicts human emotion ratings and aligns with amygdala activity.

This study shows that deep neural networks trained on visual object recognition develop emotion-selective neurons, supporting the idea that the visual system can intrinsically represent affective significance.