1.School of Biomedical Engineering and Technology, Tianjin Medical University, Tianjin 300070, China 2.Cangzhou People's Hospital, Cangzhou 061000, China
Fund Project:Project supported by the National Natural Science Foundation of China (Grant Nos. 61473208, 61876132), and Tianjin Research Program of Application Foundation and Advanced Technology (Grant No. 15JCYBJC47700).
Received Date:20 December 2018
Accepted Date:28 January 2019
Available Online:23 March 2019
Published Online:05 April 2019
Abstract:Neural firing rate homeostasis, as an important feature of neural electrical activity, means that the firing rate in brain is maintained in a relatively stable state, and fluctuates around a constant value. Extensive experimental studies have revealed that the firing rate homeostasis is ubiquitous in brain, and provides a base for neural information processing and maintaining normal neurological functions, so that the research on neural firing rate homeostasis is a central problem in the field of neuroscience. Cortical neural network is a highly complex dynamic system with a large number of input disturbance signals and parameter perturbations due to dynamic connection. However, it remains to be further investigated how firing rate homeostasis is established in cortical neural network, furthermore, maintains robustness to these disturbances and perturbations. The feedback neural circuit with recurrent excitatory and inhibitory connection is a typical connective pattern in cortical cortex, and inhibitory synaptic plasticity plays a crucial role in achieving neural firing rate homeostasis. Here, by constructing a feedback neural network with inhibitory spike timing-dependent plasticity (STDP), we conduct a computational research to elucidate the mechanism of neural firing rate homeostasis. The results indicate that the neuronal firing rate can track the target firing rate accurately under the regulation of inhibitory synaptic plasticity, thus achieve firing rate homeostasis. In the face of external disturbances and parameter perturbations, the neuron firing rate deviates transiently from the target firing rate value, and converges to the target firing rate value at a steady state, which demonstrates that the firing rate homeostasis established by the inhibitory synaptic plasticity can maintain strong robustness. Furthermore, the analytical research qualitatively explains the firing rate homeostasis mechanism underlined by inhibitory synaptic plasticity. Finally, the simulations further demonstrate that the learning rate value and the firing rate set point value also exert a quantitative influence on the firing rate homeostasis. Overall, these findings not only gain an insight into the firing rate homeostasis mechanism underlined by inhibitory synaptic plasticity, but also inspire testable hypotheses for future experimental studies. Keywords:inhibitory synaptic plasticity/ firing rate homeostasis/ robustness
表4抑制性突触可塑性的参数取值 Table4.Parameters values of inhibitory synaptic plasticity.
-->
3.1.放电率自稳态
神经元目标放电率为5, 10, 20, 50 Hz四种不同取值时的神经元膜电位及平均放电率分别如图3—图6所示. 在这四种不同情况下, 图(a)为目标神经元膜电位图; 图(b)为采用滑动平均窗方法确定的神经元平均放电率, 其中滑动时间窗取10 ms; 图(c)为随机取一个突触前神经元, 其与目标神经元之间抑制性突触权重(下文简称抑制性突触权重)的变化曲线. 图 3 目标放电率为5 Hz时的神经元放电率自稳态分析图 (a)神经元膜电位; (b)神经元平均放电率曲线; (c)抑制性突触权重变化曲线 Figure3. Firing rate homeostasis with the target firing rate equal to 5 Hz: (a) Neural membrane potential; (b) the average firing rate; (c) the strength of inhibitory synapse.
图 4 目标放电率为10 Hz时的神经元放电率自稳态分析图 (a)神经元膜电位; (b)神经元平均放电率曲线; (c)抑制性突触权重变化曲线 Figure4. Firing rate homeostasis with the target firing rate equal to 10 Hz: (a) Neuronal membrane potential; (b) the average firing rate; (c) the strength of inhibitory synapse.
图 5 目标放电率为20 Hz时的神经元放电率自稳态分析图 (a)神经元膜电位; (b)神经元平均放电率曲线; (c)抑制性突触权重变化曲线 Figure5. Firing rate homeostasis with the target firing rate equal to 20 Hz: (a) Neuronal membrane potential; (b) the average firing rate; (c) the strength of inhibitory synapse.
图 6 目标放电率为50 Hz时的神经元放电率自稳态分析图 (a)神经元膜电位; (b)神经元平均放电率曲线; (c)抑制性突触权重变化曲线 Figure6. Firing rate homeostasis with the target firing rate equal to 50 Hz: (a) Neuronal membrane potential; (b) the average firing rate; (c) the strength of inhibitory synapse.
由以上仿真结果可以看出, 神经元在抑制性突触可塑性的调节下实际放电率稳定在设定的目标放电率, 实现放电率自稳态. 本小节进一步定量分析抑制性突触可塑性中的学习率对神经元放电率自稳态的影响. 图 11 学习率对放电率自稳态的影响 (a)参数摄动信号; (b)不同学习率时的神经元平均放电率曲线 Figure11. The effect of learning rate on neural firing rate homeostasis: (a) The parameter perturbation signal; (b) the firing rate with different learning rates.