Circuit decoupling with surrogate compensation for baseline tuning in biophysically detailed neural models
The baseline regime of a neural circuit shapes its dynamical behavior. Establishing an appropriate baseline is therefore an important part of tuning a computational circuit model. This baseline can be specified in terms of target population firing rates and achieved by adjusting the background inputs to each population. Due to recurrent coupling, these inputs need to be tuned jointly, requiring a…
The baseline state of a neural network determines its behavior. Properly setting this baseline is crucial when creating computational models. This baseline can be defined using desired firing rates for groups of neurons, accomplished by tweaking the background inputs to each group. However, due to connections among groups, tuning these inputs simultaneously is a complex task, particularly for detailed models of brain circuits. This issue has limited the effectiveness of current optimization techniques.
To address this challenge, researchers created a new method that separates the circuit into individual units, enabling independent tuning of background inputs. However, when the connections are re-established, the solution might not hold. To mitigate this, the team used surrogate inputs – simulated neurons firing at the desired rates – to replace the missing recurrent connections. This approach ensures the background inputs, once fine-tuned, can still work after the connections are restored.
The researchers tested their method on a model of the macaque auditory thalamocortical circuit, which has been difficult to fine-tune. First, they adjusted the background inputs for each group of neurons individually while the circuit was fully disconnected. Then, they examined the solution within isolated groups while keeping their connections intact.
In the cortical group, they found two competing states, with the model defaulting to the less desirable one. By gradually transitioning from using surrogate neurons to relying on their own excitations, the model was able to reach a desired state.
The thalamus, on the other hand, showed a single state that didn't match the target due to imperfect surrogate inputs. However, adding a separate adjustment for each thalamus group significantly improved the results. Finally, they tested the entire circuit with the tuned background inputs, adjusted thalamus groups, and the gradual reintroduction of self-excitement.
Across multiple simulations, the full circuit consistently reached a state close to the target. This improved model of the auditory thalamocortical system can serve as a useful resource for the scientific community to explore the dynamics of this brain region.
Rewritten by urgent.news; may contain errors — check the original.
Source: biorxiv.org