Limits...
A computational model for preplay in the hippocampus.

Azizi AH, Wiskott L, Cheng S - Front Comput Neurosci (2013)

Bottom Line: Recent experiments found this correlation even between offline sequential activity (OSA) recorded before the animal ran in a novel environment and the place fields in that environment.Our results suggest two different accounts for preplay.Either an existing chart is re-used to represent a novel environment or a new chart is formed.

View Article: PubMed Central - PubMed

Affiliation: Mercator Research Group "Structure of Memory," Department of Psychology, Ruhr-University Bochum Bochum, Germany.

ABSTRACT
The hippocampal network produces sequences of neural activity even when there is no time-varying external drive. In offline states, the temporal sequence in which place cells fire spikes correlates with the sequence of their place fields. Recent experiments found this correlation even between offline sequential activity (OSA) recorded before the animal ran in a novel environment and the place fields in that environment. This preplay phenomenon suggests that OSA is generated intrinsically in the hippocampal network, and not established by external sensory inputs. Previous studies showed that continuous attractor networks with asymmetric patterns of connectivity, or with slow, local negative feedback, can generate sequential activity. This mechanism could account for preplay if the network only represented a single spatial map, or chart. However, global remapping in the hippocampus implies that multiple charts are represented simultaneously in the hippocampal network and it remains unknown whether the network with multiple charts can account for preplay. Here we show that it can. Driven with random inputs, the model generates sequences in every chart. Place fields in a given chart and OSA generated by the network are highly correlated. We also find significant correlations, albeit less frequently, even when the OSA is correlated with a new chart in which place fields are randomly scattered. These correlations arise from random correlations between the orderings of place fields in the new chart and those in a pre-existing chart. Our results suggest two different accounts for preplay. Either an existing chart is re-used to represent a novel environment or a new chart is formed.

No MeSH data available.


Capacity of the network. (A) For a given chart number the activity was initialized as a bump in each chart separately by providing bias input to nearby cells for 200 ms. The maximum of the standard deviation across charts is color coded for every time point. There is a sudden transition in this value between five to six charts, indicating a network capacity of five charts. (B) The same as part A, but the minimum of the standard deviation is considered. The transition occurs when the network cannot sustain the stable bump in at least one chart. Here between seven and eight charts. (C) The scaling of the storage capacity with the size of the network. The solid line indicates the line of best fit, 0.0028N − 0.14. (D) Distribution of the skewness of the distribution of average firing rates for networks with different numbers of stored charts. The lines represent cumulative fractions. See the text for an interpretation of these distributions.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3824291&req=5

Figure 10: Capacity of the network. (A) For a given chart number the activity was initialized as a bump in each chart separately by providing bias input to nearby cells for 200 ms. The maximum of the standard deviation across charts is color coded for every time point. There is a sudden transition in this value between five to six charts, indicating a network capacity of five charts. (B) The same as part A, but the minimum of the standard deviation is considered. The transition occurs when the network cannot sustain the stable bump in at least one chart. Here between seven and eight charts. (C) The scaling of the storage capacity with the size of the network. The solid line indicates the line of best fit, 0.0028N − 0.14. (D) Distribution of the skewness of the distribution of average firing rates for networks with different numbers of stored charts. The lines represent cumulative fractions. See the text for an interpretation of these distributions.

Mentions: Samsonovich and McNaughton estimated the number of stable charts in a network similar to ours to be 0.004N, where N is the number of cells in the network (Samsonovich and McNaughton, 1997). An independent, analytical estimate arrived at a similar capacity for a fully connected network (Battaglia and Treves, 1998). However, these estimates were obtained for networks without SFA and SFA can dramatically affect the activity of the network. To estimate the capacity of our network, we initiated the activity in each chart separately and observed how the activity dispersed in the chart over a 5 s period. The activity was initiated by a constant bias input Ibias = 1.92 to a clustered sub-population in a given chart. After 400 ms, the initiation input current was replaced by a constant bias current to all neurons in the network. The standard deviation of activity in the chart, in which the activity was initiated, was recorded for 5 s in 40 ms time bins. We define that a chart can support a bump if the standard deviation of the activity remains below 30 cm. Since the network can only be said to have a certain capacity, if all charts in the network can support a bump, we look for the maximum standard deviation across all charts in each time bin (Figure 10A). For our network of 2000 neurons, the capacity appears to be five (Figure 10A), which is lower than the eight predicted by Samsonovich and McNaughton's estimate, as expected. We next calculated the capacity for networks of different sizes (Figure 10C). The capacity scales linearly with the number of cells in the network 0.0028N, which is similar to the result reported by Samsonovich and McNaughton (1997), 0.004N, for a similar network without adaptation. It is surprising that the presence of adaptation that destabilizes the bump attractor only slightly reduces the coefficient without changing the qualitative relationship.


A computational model for preplay in the hippocampus.

Azizi AH, Wiskott L, Cheng S - Front Comput Neurosci (2013)

Capacity of the network. (A) For a given chart number the activity was initialized as a bump in each chart separately by providing bias input to nearby cells for 200 ms. The maximum of the standard deviation across charts is color coded for every time point. There is a sudden transition in this value between five to six charts, indicating a network capacity of five charts. (B) The same as part A, but the minimum of the standard deviation is considered. The transition occurs when the network cannot sustain the stable bump in at least one chart. Here between seven and eight charts. (C) The scaling of the storage capacity with the size of the network. The solid line indicates the line of best fit, 0.0028N − 0.14. (D) Distribution of the skewness of the distribution of average firing rates for networks with different numbers of stored charts. The lines represent cumulative fractions. See the text for an interpretation of these distributions.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3824291&req=5

Figure 10: Capacity of the network. (A) For a given chart number the activity was initialized as a bump in each chart separately by providing bias input to nearby cells for 200 ms. The maximum of the standard deviation across charts is color coded for every time point. There is a sudden transition in this value between five to six charts, indicating a network capacity of five charts. (B) The same as part A, but the minimum of the standard deviation is considered. The transition occurs when the network cannot sustain the stable bump in at least one chart. Here between seven and eight charts. (C) The scaling of the storage capacity with the size of the network. The solid line indicates the line of best fit, 0.0028N − 0.14. (D) Distribution of the skewness of the distribution of average firing rates for networks with different numbers of stored charts. The lines represent cumulative fractions. See the text for an interpretation of these distributions.
Mentions: Samsonovich and McNaughton estimated the number of stable charts in a network similar to ours to be 0.004N, where N is the number of cells in the network (Samsonovich and McNaughton, 1997). An independent, analytical estimate arrived at a similar capacity for a fully connected network (Battaglia and Treves, 1998). However, these estimates were obtained for networks without SFA and SFA can dramatically affect the activity of the network. To estimate the capacity of our network, we initiated the activity in each chart separately and observed how the activity dispersed in the chart over a 5 s period. The activity was initiated by a constant bias input Ibias = 1.92 to a clustered sub-population in a given chart. After 400 ms, the initiation input current was replaced by a constant bias current to all neurons in the network. The standard deviation of activity in the chart, in which the activity was initiated, was recorded for 5 s in 40 ms time bins. We define that a chart can support a bump if the standard deviation of the activity remains below 30 cm. Since the network can only be said to have a certain capacity, if all charts in the network can support a bump, we look for the maximum standard deviation across all charts in each time bin (Figure 10A). For our network of 2000 neurons, the capacity appears to be five (Figure 10A), which is lower than the eight predicted by Samsonovich and McNaughton's estimate, as expected. We next calculated the capacity for networks of different sizes (Figure 10C). The capacity scales linearly with the number of cells in the network 0.0028N, which is similar to the result reported by Samsonovich and McNaughton (1997), 0.004N, for a similar network without adaptation. It is surprising that the presence of adaptation that destabilizes the bump attractor only slightly reduces the coefficient without changing the qualitative relationship.

Bottom Line: Recent experiments found this correlation even between offline sequential activity (OSA) recorded before the animal ran in a novel environment and the place fields in that environment.Our results suggest two different accounts for preplay.Either an existing chart is re-used to represent a novel environment or a new chart is formed.

View Article: PubMed Central - PubMed

Affiliation: Mercator Research Group "Structure of Memory," Department of Psychology, Ruhr-University Bochum Bochum, Germany.

ABSTRACT
The hippocampal network produces sequences of neural activity even when there is no time-varying external drive. In offline states, the temporal sequence in which place cells fire spikes correlates with the sequence of their place fields. Recent experiments found this correlation even between offline sequential activity (OSA) recorded before the animal ran in a novel environment and the place fields in that environment. This preplay phenomenon suggests that OSA is generated intrinsically in the hippocampal network, and not established by external sensory inputs. Previous studies showed that continuous attractor networks with asymmetric patterns of connectivity, or with slow, local negative feedback, can generate sequential activity. This mechanism could account for preplay if the network only represented a single spatial map, or chart. However, global remapping in the hippocampus implies that multiple charts are represented simultaneously in the hippocampal network and it remains unknown whether the network with multiple charts can account for preplay. Here we show that it can. Driven with random inputs, the model generates sequences in every chart. Place fields in a given chart and OSA generated by the network are highly correlated. We also find significant correlations, albeit less frequently, even when the OSA is correlated with a new chart in which place fields are randomly scattered. These correlations arise from random correlations between the orderings of place fields in the new chart and those in a pre-existing chart. Our results suggest two different accounts for preplay. Either an existing chart is re-used to represent a novel environment or a new chart is formed.

No MeSH data available.