Spontaneous generation of face recognition in untrained deep neural networks
Baek, S.; Song, M.; Jang, J.; Kim, G.; Paik, S.-B.
Show abstract
Face-selective neurons are observed in the primate visual pathway and are considered the basis of facial recognition in the brain. However, it is debated whether this neuronal selectivity can arise spontaneously, or requires training from visual experience. Here, we show that face-selective neurons arise spontaneously in random feedforward networks in the absence of learning. Using biologically inspired deep neural networks, we found that face-selective neurons arise under three different network conditions: one trained using non-face natural images, one randomized after being trained, and one never trained. We confirmed that spontaneously emerged face-selective neurons show the biological view-point-invariant characteristics observed in monkeys. Such neurons suddenly vanished when feedforward weight variation declined to a certain level. Our results suggest that innate face-selectivity originates from statistical variation of the feedforward projections in hierarchical neural networks.
Matching journals
The top 5 journals account for 50% of the predicted probability mass.
Similar papers in this journal
Similar papers in this journal
- Functional harmonics reveal multi-dimensional basis functions underlying cortical organization 94%
- In V4, pair-wise synchony and correlations depend on the structure of the population code 94%
- Spherical code of retinal orientation-selectivity enables decoding in ensembled and retinotopic operation 94%
Similar papers in this journal
Similar papers in this journal
- Manifold learning analysis suggests novel strategies for aligning single-cell multi-modalities and revealing functional genomics for neuronal electrophysiology 95%
- Real-world size of objects serves as an axis of object space 94%
- Encoding of facial features by single neurons in the human amygdala and hippocampus 93%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.