Post reply

Reply to the message by David Turell

Name:

E-mail: (optional, won't be displayed directly)

Leave this field empty:

Homepage: (optional)

Leave this field empty:

Location: (optional)

Remember me (cookie)

Category:

Subject:

Formatting help

skip to input

format text bold: [b]bold text[/b]
format text italic: [i]italic text[/i]
insert hyperlink: [link=http://example.com/]link text[/link] / [link]http://example.com/[/link]
set text color: [color=#rgb]colored text[/color]
font size: [size=small]small text[/size]; [size=large]large text[/size]
insert list: [list][*]list item[/list]
insert image: [img]http://example.com/image.jpg[/img]; left: [img=left]http://example.com/image.jpg[/img]; right: [img=right]http://example.com/image.jpg[/img]; thumbnail: [img=thumbnail]http://example.com/image.jpg[/img]; thumbnail left: [img=thumbnail-left]http://example.com/image.jpg[/img]; thumbnail right: [img=thumbnail-right]http://example.com/image.jpg[/img]

:-)
;-)
:-P
:-D
:-|
:-(

Message:

> In a mouse visual cortex:
> 
> https://www.cell.com/neuron/fulltext/S0896-6273(22)00905-9?dgcid=raven_jbs_aip_email
> 
> Summary
> "Adaptive sensory behavior is thought to depend on processing in recurrent cortical circuits, but how dynamics in these circuits shapes the integration and transmission of sensory information is not well understood. Here, we study neural coding in recurrently connected networks of neurons driven by sensory input. We show analytically how information available in the network output varies with the alignment between feedforward input and the integrating modes of the circuit dynamics. In light of this theory, we analyzed neural population activity in the visual cortex of mice that learned to discriminate visual features. We found that over learning, slow patterns of network dynamics realigned to better integrate input relevant to the discrimination task. This realignment of network dynamics could be explained by changes in excitatory-inhibitory connectivity among neurons tuned to relevant features. These results suggest that learning tunes the temporal dynamics of cortical circuits to optimally integrate relevant sensory input."
> 
> ***
> 
> "Here, we ask whether improvements in stimulus decodability over learning could arise through selective temporal integration of relevant feedforward sensory input. We first show analytically how the output of a network can be tuned to optimally discriminate pairs of input stimuli by matching its recurrent dynamics to their sensory input statistics. In particular, we show that a stimulus decoder applied to network output performs best if the dimension of network input with greatest SNR [signal to noise ratio] activates a pattern of recurrent network dynamics that decays slowly. We then study how the dynamical properties of neural circuits in mouse V1 change as animals learn to discriminate visual stimuli. Using a dynamical systems model fit to experimental data (Khan et al., 2018), we find that slowly decaying patterns in the recurrent dynamics became better aligned with high-SNR sensory input over learning. Finally, we analyze circuit models with excitatory and inhibitory neurons to explore how this alignment might arise through changes in the circuit. We find that stimulus-specific changes in connectivity between excitatory and inhibitory neurons increase the alignment of recurrent dynamics with sensory input as observed experimentally. These connectivity changes predict changes in stimulus tuning and cell type-specific reorganization of dynamics within the model, which we find to be recapitulated in the experimental data. Our findings suggest a critical role for cortical dynamics in selective temporal integration of relevant sensory information."
> 
> Comment: a highly technical article which shows how we are beginning to understand brain plasticity. To wit:  "our approach provides a unifying formalism which links statistical properties of evidence integration and population coding to the dynamical properties of the underlying recurrent network. Although we have focused on changes in network dynamics over learning, the mechanism of dynamical alignment may also provide a substrate for contextual or attentional modulation of sensory processing (Gilbert, and Li, 2013). Specifically, top-down input may modulate the dynamics of recipient neural populations, transiently aligning dynamical modes of the local circuit with relevant features of bottom-up sensory input according to task context. Such a mechanism could allow for flexible routing and gating of information between brain areas through the dynamical formation and coordination of “communication subspaces” (Semedo et al., 2019; Kohn et al., 2020; Javadzadeh and Hofer, 2022), configured through selective alignment of local modes across anatomically distributed circuits."