Presentation Information
[POS-11]Effect of Dopamine on Spatio-Temporal Spike Pattern Detection Ability of Single Neurons
*Ayaka Kotajima1, Shunta Furuichi2, Takashi Kohno2 (1. Tokyo Metropolitan Shinjyuku Yamabuki High School (Japan), 2. The University of Tokyo (Japan))
Keywords:
Computational Neuroscience,Dopamine,Spike Timing-Dependent Plasticity (STDP),Spiking Neural Network
This study focuses on spatio-temporal spike pattern detection ability of single neurons, the most fundamental units of the brain. We investigated the role of dopamine in information processing through a computational approach.
Based on the relationship between spike-timing-dependent plasticity (STDP) and dopamine demonstrated by Zhang et al., we modeled STDP learning curves with and without dopamine. The spatio-temporal spike pattern detection model for single neurons composed of the Leaky Integrate-and-Fire neuron and alpha function-type synapse models (Masquelier et al., 2008) was extended to incorporate dopamine input. By switching the STDP learning curve depending on the presence or absence of dopamine, we analyzed the effects of dopamine on learning. Dopamine was released only during specific 10 ms intervals within the 50 ms spike patterns to be detected. The effects of dopamine were evaluated concerning its release timing, release probability, initial synaptic weight, and learning duration.
As a result, the detection success rate improved by 20–40% dopamine release, but further increases led to a decline in the success rate due to increased false alarms. The difference in learning efficiency between the presence and absence of dopamine was decreased for longer learning durations. Additionally, the neuron exhibited a more pronounced effect from dopamine when the initial synaptic weights were lower. Moreover, when dopamine was released in the first 10 ms of the spike pattern, the false alarms increased significantly, leading to a marked drop in success rate. This is because the preceding spikes before the spike pattern to be detected are random thus typically long-term depression (LTD) occurs during this interval, but in the presence of dopamine, even the random spikes evoke long-term potentiation (LTP), which results in the strengthening of the association with random patterns.
This study demonstrated that dopamine enhanced learning efficiency and improved information detection capabilities. Also, it was theoretically shown that neurons with lower excitability are more strongly affected by dopamine. Furthermore, it was suggested that the timing of dopamine release has a major impact on learning efficiency, dopamine should be released with a short delay to the patterns to be detected.
Based on the relationship between spike-timing-dependent plasticity (STDP) and dopamine demonstrated by Zhang et al., we modeled STDP learning curves with and without dopamine. The spatio-temporal spike pattern detection model for single neurons composed of the Leaky Integrate-and-Fire neuron and alpha function-type synapse models (Masquelier et al., 2008) was extended to incorporate dopamine input. By switching the STDP learning curve depending on the presence or absence of dopamine, we analyzed the effects of dopamine on learning. Dopamine was released only during specific 10 ms intervals within the 50 ms spike patterns to be detected. The effects of dopamine were evaluated concerning its release timing, release probability, initial synaptic weight, and learning duration.
As a result, the detection success rate improved by 20–40% dopamine release, but further increases led to a decline in the success rate due to increased false alarms. The difference in learning efficiency between the presence and absence of dopamine was decreased for longer learning durations. Additionally, the neuron exhibited a more pronounced effect from dopamine when the initial synaptic weights were lower. Moreover, when dopamine was released in the first 10 ms of the spike pattern, the false alarms increased significantly, leading to a marked drop in success rate. This is because the preceding spikes before the spike pattern to be detected are random thus typically long-term depression (LTD) occurs during this interval, but in the presence of dopamine, even the random spikes evoke long-term potentiation (LTP), which results in the strengthening of the association with random patterns.
This study demonstrated that dopamine enhanced learning efficiency and improved information detection capabilities. Also, it was theoretically shown that neurons with lower excitability are more strongly affected by dopamine. Furthermore, it was suggested that the timing of dopamine release has a major impact on learning efficiency, dopamine should be released with a short delay to the patterns to be detected.