Limits...
Goal-directed control with cortical units that are gated by both top-down feedback and oscillatory coherence.

Kerr RR, Grayden DB, Thomas DA, Gilson M, Burkitt AN - Front Neural Circuits (2014)

Bottom Line: We demonstrate that more sophisticated and flexible top-down control is possible when the gain of units is modulated by not only top-down feedback but by coherence between the activities of the oscillating units.With these types of units, it is possible to not only add units to, or remove units from, a higher-level unit's logic operation using top-down feedback, but also to modify the type of role that a unit plays in the operation.Based on this, we make predictions about the likely connectivities between certain brain regions that have been experimentally observed to be involved in goal-directed behavior and top-down attention.

View Article: PubMed Central - PubMed

Affiliation: NeuroEngineering Laboratory, Department of Electrical and Electronic Engineering, The University of Melbourne Melbourne, VIC, Australia ; Centre for Neural Engineering, The University of Melbourne Melbourne, VIC, Australia ; NICTA, Victoria Research Lab, The University of Melbourne Melbourne, VIC, Australia.

ABSTRACT
The brain is able to flexibly select behaviors that adapt to both its environment and its present goals. This cognitive control is understood to occur within the hierarchy of the cortex and relies strongly on the prefrontal and premotor cortices, which sit at the top of this hierarchy. Pyramidal neurons, the principal neurons in the cortex, have been observed to exhibit much stronger responses when they receive inputs at their soma/basal dendrites that are coincident with inputs at their apical dendrites. This corresponds to inputs from both lower-order regions (feedforward) and higher-order regions (feedback), respectively. In addition to this, coherence between oscillations, such as gamma oscillations, in different neuronal groups has been proposed to modulate and route communication in the brain. In this paper, we develop a simple, but novel, neural mass model in which cortical units (or ensembles) exhibit gamma oscillations when they receive coherent oscillatory inputs from both feedforward and feedback connections. By forming these units into circuits that can perform logic operations, we identify the different ways in which operations can be initiated and manipulated by top-down feedback. We demonstrate that more sophisticated and flexible top-down control is possible when the gain of units is modulated by not only top-down feedback but by coherence between the activities of the oscillating units. With these types of units, it is possible to not only add units to, or remove units from, a higher-level unit's logic operation using top-down feedback, but also to modify the type of role that a unit plays in the operation. Finally, we explore how different network properties affect top-down control and processing in large networks. Based on this, we make predictions about the likely connectivities between certain brain regions that have been experimentally observed to be involved in goal-directed behavior and top-down attention.

Show MeSH
Robustness to noisy oscillations. (A) Performance, given by the fraction of trials in which the output unit was activated, of the “OR” network shown in Figure 3 (with both inputs active) to different amounts and types of noise, θ. These different types are: noise only for feedback inputs (blue), noise only for feedforward inputs (red), and noise for both feedforward and feedback inputs (green). The noise can be either “simple noise” (solid) or “peak-only noise” (dashed). Markers show the average outcome from 1000 simulations and lines show analytically derived curves. (B) Same as (A) but where one input is active and one is inactive. (C) Same as (A) but for the “AND NOT” network shown in Figure 3 where one input is active and one is inactive. (D) Same as (A) but for the “AND” network shown in Figure 3 where both inputs are active and without the analytically derived lines.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4126059&req=5

Figure 9: Robustness to noisy oscillations. (A) Performance, given by the fraction of trials in which the output unit was activated, of the “OR” network shown in Figure 3 (with both inputs active) to different amounts and types of noise, θ. These different types are: noise only for feedback inputs (blue), noise only for feedforward inputs (red), and noise for both feedforward and feedback inputs (green). The noise can be either “simple noise” (solid) or “peak-only noise” (dashed). Markers show the average outcome from 1000 simulations and lines show analytically derived curves. (B) Same as (A) but where one input is active and one is inactive. (C) Same as (A) but for the “AND NOT” network shown in Figure 3 where one input is active and one is inactive. (D) Same as (A) but for the “AND” network shown in Figure 3 where both inputs are active and without the analytically derived lines.

Mentions: In Figures 9A,B, we explore the robustness to noise of the “OR” network when both input units receive feedforward activity and when only one input unit receives feedforward activity, respectively. In Figures 9C,D, we explore the robustness to noise of the “AND NOT” and “AND” networks, respectively, when both input units receive feedforward activity. We plot the fraction of trials (out of 1000) in which the output units in each of these networks were activated for different amounts and types of noise (shown by markers). For each of these networks (except for the more complex “AND” network), we compared these simulation results to analytically derived functions for the probability of activation of the output units (shown by lines). For “simple noise,” when θ = 0.5, no oscillations are present (i.e., the likelihood of activity in any time-step is the same) and very poor performance is observed. This highlights the importance oscillations play in these networks. For noise levels of θ = 0.1, the performance dropped down to as low as 40%. However, we observed that “simple noise” impairs the performance of the networks much more than “peak-only noise” in each of the different networks. “Peak-only noise” where θ < 0.05 was sufficient to ensure activation at least 80% of the time (shown by black, dashed lines).


Goal-directed control with cortical units that are gated by both top-down feedback and oscillatory coherence.

Kerr RR, Grayden DB, Thomas DA, Gilson M, Burkitt AN - Front Neural Circuits (2014)

Robustness to noisy oscillations. (A) Performance, given by the fraction of trials in which the output unit was activated, of the “OR” network shown in Figure 3 (with both inputs active) to different amounts and types of noise, θ. These different types are: noise only for feedback inputs (blue), noise only for feedforward inputs (red), and noise for both feedforward and feedback inputs (green). The noise can be either “simple noise” (solid) or “peak-only noise” (dashed). Markers show the average outcome from 1000 simulations and lines show analytically derived curves. (B) Same as (A) but where one input is active and one is inactive. (C) Same as (A) but for the “AND NOT” network shown in Figure 3 where one input is active and one is inactive. (D) Same as (A) but for the “AND” network shown in Figure 3 where both inputs are active and without the analytically derived lines.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4126059&req=5

Figure 9: Robustness to noisy oscillations. (A) Performance, given by the fraction of trials in which the output unit was activated, of the “OR” network shown in Figure 3 (with both inputs active) to different amounts and types of noise, θ. These different types are: noise only for feedback inputs (blue), noise only for feedforward inputs (red), and noise for both feedforward and feedback inputs (green). The noise can be either “simple noise” (solid) or “peak-only noise” (dashed). Markers show the average outcome from 1000 simulations and lines show analytically derived curves. (B) Same as (A) but where one input is active and one is inactive. (C) Same as (A) but for the “AND NOT” network shown in Figure 3 where one input is active and one is inactive. (D) Same as (A) but for the “AND” network shown in Figure 3 where both inputs are active and without the analytically derived lines.
Mentions: In Figures 9A,B, we explore the robustness to noise of the “OR” network when both input units receive feedforward activity and when only one input unit receives feedforward activity, respectively. In Figures 9C,D, we explore the robustness to noise of the “AND NOT” and “AND” networks, respectively, when both input units receive feedforward activity. We plot the fraction of trials (out of 1000) in which the output units in each of these networks were activated for different amounts and types of noise (shown by markers). For each of these networks (except for the more complex “AND” network), we compared these simulation results to analytically derived functions for the probability of activation of the output units (shown by lines). For “simple noise,” when θ = 0.5, no oscillations are present (i.e., the likelihood of activity in any time-step is the same) and very poor performance is observed. This highlights the importance oscillations play in these networks. For noise levels of θ = 0.1, the performance dropped down to as low as 40%. However, we observed that “simple noise” impairs the performance of the networks much more than “peak-only noise” in each of the different networks. “Peak-only noise” where θ < 0.05 was sufficient to ensure activation at least 80% of the time (shown by black, dashed lines).

Bottom Line: We demonstrate that more sophisticated and flexible top-down control is possible when the gain of units is modulated by not only top-down feedback but by coherence between the activities of the oscillating units.With these types of units, it is possible to not only add units to, or remove units from, a higher-level unit's logic operation using top-down feedback, but also to modify the type of role that a unit plays in the operation.Based on this, we make predictions about the likely connectivities between certain brain regions that have been experimentally observed to be involved in goal-directed behavior and top-down attention.

View Article: PubMed Central - PubMed

Affiliation: NeuroEngineering Laboratory, Department of Electrical and Electronic Engineering, The University of Melbourne Melbourne, VIC, Australia ; Centre for Neural Engineering, The University of Melbourne Melbourne, VIC, Australia ; NICTA, Victoria Research Lab, The University of Melbourne Melbourne, VIC, Australia.

ABSTRACT
The brain is able to flexibly select behaviors that adapt to both its environment and its present goals. This cognitive control is understood to occur within the hierarchy of the cortex and relies strongly on the prefrontal and premotor cortices, which sit at the top of this hierarchy. Pyramidal neurons, the principal neurons in the cortex, have been observed to exhibit much stronger responses when they receive inputs at their soma/basal dendrites that are coincident with inputs at their apical dendrites. This corresponds to inputs from both lower-order regions (feedforward) and higher-order regions (feedback), respectively. In addition to this, coherence between oscillations, such as gamma oscillations, in different neuronal groups has been proposed to modulate and route communication in the brain. In this paper, we develop a simple, but novel, neural mass model in which cortical units (or ensembles) exhibit gamma oscillations when they receive coherent oscillatory inputs from both feedforward and feedback connections. By forming these units into circuits that can perform logic operations, we identify the different ways in which operations can be initiated and manipulated by top-down feedback. We demonstrate that more sophisticated and flexible top-down control is possible when the gain of units is modulated by not only top-down feedback but by coherence between the activities of the oscillating units. With these types of units, it is possible to not only add units to, or remove units from, a higher-level unit's logic operation using top-down feedback, but also to modify the type of role that a unit plays in the operation. Finally, we explore how different network properties affect top-down control and processing in large networks. Based on this, we make predictions about the likely connectivities between certain brain regions that have been experimentally observed to be involved in goal-directed behavior and top-down attention.

Show MeSH