Stimuli Reduce the Dimensionality of Cortical Activity

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Neural activity is often analyzed in high-dimensional firing rate spaces.
• Neural activity can be effectively represented in lower-dimensional subspaces.
• Understanding neural dimensionality is key to deciphering neural computation.

Purpose of the Study:

• Investigate neural ensemble dimensionality in the sensory cortex of alert rats.
• Compare dimensionality during ongoing versus stimulus-evoked activity.
• Develop a theoretical framework for neural dimensionality.

Main Methods:

• Recorded neural ensembles from the sensory cortex of alert rats.
• Analyzed neural activity during inter-trial (ongoing) and stimulus-evoked periods.
• Utilized a spiking network model with a clustered architecture.
• Developed a theoretical model to predict dimensionality bounds.

Main Results:

• Neural dimensionality scales linearly with ensemble size.
• Dimensionality grows significantly faster during ongoing activity than evoked activity.
• A clustered network model accurately predicts observed scaling and differences between activity states.
• A theoretical upper bound on dimensionality was derived, dependent on correlations and cluster number.

Conclusions:

• Neural dimensionality is a dynamic property influenced by brain states (ongoing vs. evoked activity).
• Clustered network architecture plays a crucial role in shaping neural dimensionality.
• The findings provide a framework for analyzing and understanding neural dimensionality in vivo and in silico.