Familiarity training enhance straightening of neural trajectory for video prediction
Qiu, W.; Huang, S.; Mukundan, M.; Lee, T. S.
Show abstract
Predictive processing in the visual system is pivotal for efficient sensory-driven behaviors. Previous research has shown that the visual system transforms sequential inputs into straighter temporal trajectories. However, the specific role of neural straightening in predictive processing, especially in how learning reshapes this phenomenon for enhanced prediction, remains unclear. To address this, we analyzed V2 population activity in macaques during familiarity training with video stimuli. Our findings reveal that repeated exposure to the same movies significantly enhances neural straightening, indicating a critical role of learning in refining neural trajectories for prediction. In parallel, our studies with the deep predictive network model, Pred-Net, demonstrated similar enhancements in neural straightening in response to familiar movies. This underscores a strong association between neural straightening and predictive coding. Together, our results provide novel insights into the adaptive mechanisms of the visual cortex, enriching our understanding of how learning shapes neural path-ways for efficient prediction.
Matching journals
The top 6 journals account for 50% of the predicted probability mass.
Similar papers in this journal
Similar papers in this journal
- Contrast Sensitivity Function in Deep Networks 96%
- The role of capacity constraints in Convolutional Neural Networks for learning random versus natural data 95%
- A systematic analysis of the joint effects of ganglion cells, lagged LGN cells, and intercortical inhibition on spatiotemporal processing and direction selectivity 95%
Similar papers in this journal
- Probing the Structure and Functional Properties of the Dropout-induced Correlated Variability in Convolutional Neural Networks 96%
- Efficient decoding of large-scale neural population responses with Gaussian-process multiclass regression 95%
- Heterogeneity in Neuronal Dynamics is Learned by Gradient Descent for Temporal Processing Tasks 95%
Similar papers in this journal
Similar papers in this journal
"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.