A population vector model of visual working memory for real-world scenes.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Visual Perception

Background:

• Visual working memory (VWM) is crucial for real-world interaction.
• Current VWM models struggle with complex, naturalistic scenes, focusing on simplified artificial objects.
• A gap exists in understanding VWM for ecologically valid visual inputs.

Purpose of the Study:

• To introduce a novel population vector model for VWM of naturalistic scenes.
• To bridge the gap between simplified VWM models and the complexity of real-world visual input.
• To enhance the ecological validity of computational models for VWM.

Main Methods:

• Developed a population vector model representing scenes as neural firing rate vectors.
• Utilized deep neural networks to estimate neural representations from natural scenes.
• Tested the model's ability to explain behavioral and neural data in VWM tasks.

Main Results:

• The population vector model successfully accounts for behavioral performance variations.
• The model explains observed patterns of brain activity during naturalistic scene storage.
• Demonstrated the model's efficacy in representing complex visual information in VWM.

Conclusions:

• The proposed population vector model offers a viable approach for VWM of real-world scenes.
• This work paves the way for more sophisticated models of naturalistic scene representation in memory.
• Advances the understanding of how complex visual environments are encoded and maintained in working memory.