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
Memories induced by correlated stimuli show distinct spike timings.Learning correlated stimuli generates distinct memories with different spike timings but the same firing activity. (A) State vectors induced by the reference stimulus (c = 1) and a correlated stimulus (c = 0.9) show large different timings in a network with . (B) State vectors induced by the reference stimulus (c = 1) and a correlated stimulus (c = 0.9) show small different timings in a network with . (C) SV distance becomes significantly larger when ; (D) Correlation coefficients r nearly overlap to each other. Here  is used to compare the case with d. In (A) and (B),  means that there is only one different input cell within the stimulus consisted of total 10 input cells. Therefore, the difference between  and  is due to only one unshared input cell.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2710501&req=5

pone-0006247-g005: Memories induced by correlated stimuli show distinct spike timings.Learning correlated stimuli generates distinct memories with different spike timings but the same firing activity. (A) State vectors induced by the reference stimulus (c = 1) and a correlated stimulus (c = 0.9) show large different timings in a network with . (B) State vectors induced by the reference stimulus (c = 1) and a correlated stimulus (c = 0.9) show small different timings in a network with . (C) SV distance becomes significantly larger when ; (D) Correlation coefficients r nearly overlap to each other. Here is used to compare the case with d. In (A) and (B), means that there is only one different input cell within the stimulus consisted of total 10 input cells. Therefore, the difference between and is due to only one unshared input cell.

Mentions: Fig. 5(A) presents typical state vectors with different c in a locally connected network with . Here we showed two memory states induced by two correlated stimuli that were consisted of 9 shared and only 1 unshared input E-cells, which generated two state vectors at and . The state vector (green) of the reference state and (yellow) of the correlated stimulus were quite different in their timings. Such a large difference of memory states was resulted from the small difference in two stimuli with only one distinct input cell. Similarly, Fig. 5(B) exhibits two memory states induced by the same stimuli but in a globally connected network (). Note the timing difference of state vectors was smaller than that in Fig. 5(A), which suggested that the network topology had an effect in the learning of correlated stimuli. We provided a whole picture with the full range of the variation of stimulus correlation in Fig. 5(C,D). As shown in Fig. 5(C), increased with the decreased correlation index c, which was a consequence of the stimuli-learning-memory dependency: distinct input stimuli generated distinct STPs. The learning behavior of the globally connected network with was dramatically different from that with the local connected topology with . The distance was doubled or more with local connections, as a result, memory states were significantly different. For a correlated stimulus, the distance between its memory state and the reference state was large when , which suggested that local connections can enhance the network to clarify the specificity of a stimulus by holding distinct memory states. This may be important for the mechanisms of discrimination between correlated memory states during the learning of similar input stimuli. Back to the example at the beginning, the larger distance between correlated ‘A’ will make ‘A’ more sensible and easy to be clarified. Given the sensory system is a bounded space, the locally connected network has a large sensible space to memorize more signals. In this sense, the local connected topology is analogous to the brain ssytem with different and local topographic areas.


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

Liu JK, She ZS - PLoS ONE (2009)

Memories induced by correlated stimuli show distinct spike timings.Learning correlated stimuli generates distinct memories with different spike timings but the same firing activity. (A) State vectors induced by the reference stimulus (c = 1) and a correlated stimulus (c = 0.9) show large different timings in a network with . (B) State vectors induced by the reference stimulus (c = 1) and a correlated stimulus (c = 0.9) show small different timings in a network with . (C) SV distance becomes significantly larger when ; (D) Correlation coefficients r nearly overlap to each other. Here  is used to compare the case with d. In (A) and (B),  means that there is only one different input cell within the stimulus consisted of total 10 input cells. Therefore, the difference between  and  is due to only one unshared input cell.
© Copyright Policy
Related In: Results  -  Collection

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

pone-0006247-g005: Memories induced by correlated stimuli show distinct spike timings.Learning correlated stimuli generates distinct memories with different spike timings but the same firing activity. (A) State vectors induced by the reference stimulus (c = 1) and a correlated stimulus (c = 0.9) show large different timings in a network with . (B) State vectors induced by the reference stimulus (c = 1) and a correlated stimulus (c = 0.9) show small different timings in a network with . (C) SV distance becomes significantly larger when ; (D) Correlation coefficients r nearly overlap to each other. Here is used to compare the case with d. In (A) and (B), means that there is only one different input cell within the stimulus consisted of total 10 input cells. Therefore, the difference between and is due to only one unshared input cell.
Mentions: Fig. 5(A) presents typical state vectors with different c in a locally connected network with . Here we showed two memory states induced by two correlated stimuli that were consisted of 9 shared and only 1 unshared input E-cells, which generated two state vectors at and . The state vector (green) of the reference state and (yellow) of the correlated stimulus were quite different in their timings. Such a large difference of memory states was resulted from the small difference in two stimuli with only one distinct input cell. Similarly, Fig. 5(B) exhibits two memory states induced by the same stimuli but in a globally connected network (). Note the timing difference of state vectors was smaller than that in Fig. 5(A), which suggested that the network topology had an effect in the learning of correlated stimuli. We provided a whole picture with the full range of the variation of stimulus correlation in Fig. 5(C,D). As shown in Fig. 5(C), increased with the decreased correlation index c, which was a consequence of the stimuli-learning-memory dependency: distinct input stimuli generated distinct STPs. The learning behavior of the globally connected network with was dramatically different from that with the local connected topology with . The distance was doubled or more with local connections, as a result, memory states were significantly different. For a correlated stimulus, the distance between its memory state and the reference state was large when , which suggested that local connections can enhance the network to clarify the specificity of a stimulus by holding distinct memory states. This may be important for the mechanisms of discrimination between correlated memory states during the learning of similar input stimuli. Back to the example at the beginning, the larger distance between correlated ‘A’ will make ‘A’ more sensible and easy to be clarified. Given the sensory system is a bounded space, the locally connected network has a large sensible space to memorize more signals. In this sense, the local connected topology is analogous to the brain ssytem with different and local topographic areas.

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