Intrinsic physiology of inhibitory neurons changes over auditory development

Unsupervised Spiking Neural Network with Dynamic Learning of Inhibitory Neurons

A spiking neural network (SNN) is a type of artificial neural network that operates based on discrete spikes to process timing information, similar to the manner in which the human brain processes real-world problems. In this paper, we propose a new spiking neural network (SNN) based on conventional, biologically plausible paradigms, such as the leaky integrate-and-fire model, spike timing-dependent plasticity, and the adaptive spiking threshold, by suggesting new biological models; that is, dynamic inhibition weight change, a synaptic wiring method, and Bayesian inference. The proposed network is designed for image recognition tasks, which are frequently used to evaluate the performance of conventional deep neural networks. To manifest the bio-realistic neural architecture, the learning is unsupervised, and the inhibition weight is dynamically changed; this, in turn, affects the synaptic wiring method based on Hebbian learning and the neuronal population. In the inference phase, Bayesian inference successfully classifies the input digits by counting the spikes from the responding neurons. The experimental results demonstrate that the proposed biological model ensures a performance improvement compared with other biologically plausible SNN models.

Fast-slow analysis as a technique for understanding the neuronal response to current ramps

The standard protocol for studying the spiking properties of single neurons is the application of current steps while monitoring the voltage response. Although this is informative, the jump in applied current is artificial. A more physiological input is where the applied current is ramped up, reflecting chemosensory input. Unsurprisingly, neurons can respond differently to the two protocols, since ion channel activation and inactivation are affected differently. Understanding the effects of current ramps, and changes in their slopes, is facilitated by mathematical models. However, techniques for analyzing current ramps are under-developed. In this article, we demonstrate how current ramps can be analyzed in single neuron models. The primary issue is the presence of gating variables that activate on slow time scales and are therefore far from equilibrium throughout the ramp. The use of an appropriate fast-slow analysis technique allows one to fully understand the neural response to ramps of different slopes. This study is motivated by data from olfactory bulb dopamine neurons, where both fast ramp (tens of milliseconds) and slow ramp (tens of seconds) protocols are used to understand the spiking profiles of the cells. The slow ramps generate experimental bifurcation diagrams with the applied current as a bifurcation parameter, thereby establishing asymptotic spiking activity patterns. The faster ramps elicit purely transient behavior that is of relevance to most physiological inputs, which are short in duration. The two protocols together provide a broader understanding of the neuron's spiking profile and the role that slowly activating ion channels can play.

Intrinsic properties of avian interaural level difference sound localizing neurons

Intrinsic properties of neurons are one major determinant for how neurons respond to their synaptic inputs and shape their outputs in neural circuits. Here, we studied the intrinsic properties of neurons in the chicken posterior portion of the dorsal nucleus of the lateral lemniscus (LLDp), the first interaural level difference (ILD) encoder of the avian auditory pathway. Using whole-cell recordings in brain slices, we revealed that the LLDp is composed of a heterogeneous neuron population based on their firing patterns. LLDp neurons were broadly classified as either phasic or tonic firing neurons, with further classification applied to tonic firing neurons, such as regular (most dominant, n = 82 out of 125 cells, 65.6%), pauser, or adaptive firing. Neurons with different firing patterns were distributed about evenly across the dorsoventral as well as mediolateral axis of LLDp. Phasic firing neurons were of faster membrane time constant, and lower excitability than tonic firing neurons. The action potentials (APs) elicited at the current thresholds displayed significant differences in first spike latency, AP peak amplitude, half-width, and maximal rising and falling rates. Interestingly, for APs elicited at suprathreshold currents (400 pA above thresholds), some of the differences diminished while a few others emerged. Remarkably, most parameters of the APs at thresholds were significantly different from those of APs at suprathresholds. Combined with our previous study (Curry and Lu, 2016), the results lend support to the two-cell type model for ILD coding in the avian system.