Limits...
A spike-timing pattern based neural network model for the study of memory dynamics.

Liu JK, She ZS - PLoS ONE (2009)

Bottom Line: We show that the distance measure can capture the timing difference of memory states.In addition, we examine the influence of network topology on learning ability, and show that local connections can increase the network's ability to embed more memory states.Together theses results suggest that the proposed system based on spike-timing patterns gives a productive model for the study of detailed learning and memory dynamics.

View Article: PubMed Central - PubMed

Affiliation: Department of Mathematics, University of California Los Angeles, Los Angeles, California, United States of America. liujk@ucla.edu

ABSTRACT
It is well accepted that the brain's computation relies on spatiotemporal activity of neural networks. In particular, there is growing evidence of the importance of continuously and precisely timed spiking activity. Therefore, it is important to characterize memory states in terms of spike-timing patterns that give both reliable memory of firing activities and precise memory of firing timings. The relationship between memory states and spike-timing patterns has been studied empirically with large-scale recording of neuron population in recent years. Here, by using a recurrent neural network model with dynamics at two time scales, we construct a dynamical memory network model which embeds both fast neural and synaptic variation and slow learning dynamics. A state vector is proposed to describe memory states in terms of spike-timing patterns of neural population, and a distance measure of state vector is defined to study several important phenomena of memory dynamics: partial memory recall, learning efficiency, learning with correlated stimuli. We show that the distance measure can capture the timing difference of memory states. In addition, we examine the influence of network topology on learning ability, and show that local connections can increase the network's ability to embed more memory states. Together theses results suggest that the proposed system based on spike-timing patterns gives a productive model for the study of detailed learning and memory dynamics.

Show MeSH
Network in response to partial cues shows unreliable memory with the distracted spike timings.(A,B) Network response measured with firing rate  (A) and distance  (B) as a function of the intensity of partial cue (E-input) and connection diameter. Both are normalized and  to compare with . (C) Overestimation of recalled memory  are the subtracted matrix (B) from (A). (D) Average  over all values of diam increases with the intensity of partial cues. Error bars (S.E.M.) are collected from 3 stimulations with different random number seeds. Each data point is an average result of the variation of the intensity of I-inputs.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2710501&req=5

pone-0006247-g004: Network in response to partial cues shows unreliable memory with the distracted spike timings.(A,B) Network response measured with firing rate (A) and distance (B) as a function of the intensity of partial cue (E-input) and connection diameter. Both are normalized and to compare with . (C) Overestimation of recalled memory are the subtracted matrix (B) from (A). (D) Average over all values of diam increases with the intensity of partial cues. Error bars (S.E.M.) are collected from 3 stimulations with different random number seeds. Each data point is an average result of the variation of the intensity of I-inputs.

Mentions: The STP described by SV was a natural characteristic of the memory state. With SV, one can characterize the memory recall process in terms of the distance between STPs instead of firing rate [20]. In general, memory recall is assumed as a process of the full memory recovery. Here it was described by the intensity of the response to a fraction of E-inputs previously memorized (partial cues), then we studied this process by varying the number of stimulated E-cells (E-inputs) after the whole stimulus was learned. This process was designed as follows. Let the network evolved first with the full cue (the whole set of stimulus) until it reached its steady state, then learning was turned off since synaptic weights were stabilized, and we increased the intensity of E-inputs stepwise as the partial cues of associative memory [20]. The changing of the number of inhibitory inputs had little effect comparing with E-inputs, thus the result below were averaged over all fractions of inhibitory inputs. Figure 4(A,B) shows the evolving of the recall function with different partial cues, where two measures of recalled memory quality were calculated,


A spike-timing pattern based neural network model for the study of memory dynamics.

Liu JK, She ZS - PLoS ONE (2009)

Network in response to partial cues shows unreliable memory with the distracted spike timings.(A,B) Network response measured with firing rate  (A) and distance  (B) as a function of the intensity of partial cue (E-input) and connection diameter. Both are normalized and  to compare with . (C) Overestimation of recalled memory  are the subtracted matrix (B) from (A). (D) Average  over all values of diam increases with the intensity of partial cues. Error bars (S.E.M.) are collected from 3 stimulations with different random number seeds. Each data point is an average result of the variation of the intensity of I-inputs.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2710501&req=5

pone-0006247-g004: Network in response to partial cues shows unreliable memory with the distracted spike timings.(A,B) Network response measured with firing rate (A) and distance (B) as a function of the intensity of partial cue (E-input) and connection diameter. Both are normalized and to compare with . (C) Overestimation of recalled memory are the subtracted matrix (B) from (A). (D) Average over all values of diam increases with the intensity of partial cues. Error bars (S.E.M.) are collected from 3 stimulations with different random number seeds. Each data point is an average result of the variation of the intensity of I-inputs.
Mentions: The STP described by SV was a natural characteristic of the memory state. With SV, one can characterize the memory recall process in terms of the distance between STPs instead of firing rate [20]. In general, memory recall is assumed as a process of the full memory recovery. Here it was described by the intensity of the response to a fraction of E-inputs previously memorized (partial cues), then we studied this process by varying the number of stimulated E-cells (E-inputs) after the whole stimulus was learned. This process was designed as follows. Let the network evolved first with the full cue (the whole set of stimulus) until it reached its steady state, then learning was turned off since synaptic weights were stabilized, and we increased the intensity of E-inputs stepwise as the partial cues of associative memory [20]. The changing of the number of inhibitory inputs had little effect comparing with E-inputs, thus the result below were averaged over all fractions of inhibitory inputs. Figure 4(A,B) shows the evolving of the recall function with different partial cues, where two measures of recalled memory quality were calculated,

Bottom Line: We show that the distance measure can capture the timing difference of memory states.In addition, we examine the influence of network topology on learning ability, and show that local connections can increase the network's ability to embed more memory states.Together theses results suggest that the proposed system based on spike-timing patterns gives a productive model for the study of detailed learning and memory dynamics.

View Article: PubMed Central - PubMed

Affiliation: Department of Mathematics, University of California Los Angeles, Los Angeles, California, United States of America. liujk@ucla.edu

ABSTRACT
It is well accepted that the brain's computation relies on spatiotemporal activity of neural networks. In particular, there is growing evidence of the importance of continuously and precisely timed spiking activity. Therefore, it is important to characterize memory states in terms of spike-timing patterns that give both reliable memory of firing activities and precise memory of firing timings. The relationship between memory states and spike-timing patterns has been studied empirically with large-scale recording of neuron population in recent years. Here, by using a recurrent neural network model with dynamics at two time scales, we construct a dynamical memory network model which embeds both fast neural and synaptic variation and slow learning dynamics. A state vector is proposed to describe memory states in terms of spike-timing patterns of neural population, and a distance measure of state vector is defined to study several important phenomena of memory dynamics: partial memory recall, learning efficiency, learning with correlated stimuli. We show that the distance measure can capture the timing difference of memory states. In addition, we examine the influence of network topology on learning ability, and show that local connections can increase the network's ability to embed more memory states. Together theses results suggest that the proposed system based on spike-timing patterns gives a productive model for the study of detailed learning and memory dynamics.

Show MeSH