| name | dgn-dynamic-gated-neuron |
| description | Dynamic Gated Neuron (DGN) — brain-inspired gating mechanism for robust spiking neural networks. Dynamic conductance acts as biologically plausible gating for selective input filtering and adaptive noise suppression. Use when designing robust SNNs, implementing noise-resilient spike-based computation, or building biologically realistic neuron models with enhanced stochastic stability. |
| user-invocable | true |
Dynamic Gated Neuron (DGN)
来源论文： arXiv:2509.03281 (2025-09-03) - "A Brain-Inspired Gating Mechanism Unlocks Robust Computation in Spiking Neural Networks"
作者: Qianyi Bai, Haiteng Wang, Qiang Yu
核心方法论
1. 核心洞察
传统 LIF 神经元省略了生物神经元固有的动态电导机制，限制了应对噪声和时间变异性的能力。动态电导本质上是生物学合理的门控机制，可以调节信息流、选择性过滤输入和自适应抑制噪声。
2. DGN 神经元模型
传统 LIF (基准)
τ_m dV/dt = -V + R * I_syn
当 V ≥ V_th: 发放脉冲, V → V_reset
DGN (改进)
τ_m dV/dt = -(V - V_rest) - g(t) * (V - E_syn) + R * I_syn(t)
dg/dt = (-g + g_0)/τ_g + β * spike_activity
其中:
- g(t): 动态膜电导，随神经活动演化
- g_0: 基线电导
- τ_g: 电导时间常数
- β: 活动-电导耦合系数
- E_syn: 突触反转电位
门控机制解释
- 高 g(t) → 强"门控" → 输入被抑制 → 噪声过滤
- 低 g(t) → 弱"门控" → 输入被传递 → 信号通过
- 电导自动响应神经活动，实现自适应门控
3. 理论保证
随机稳定性
DGN 相比标准 LIF 具有增强的随机稳定性：
定理: 在加性噪声 ξ(t) 下，
DGN 的方差 Var[V_DGN] < Var[V_LIF]
证明思路: 动态电导作为扰动抑制机制
dg/dt 项引入了负反馈，稳定膜电位波动
噪声抑制
信噪比增益:
SNR_DGN / SNR_LIF ≈ 1 + (g_dynamic / g_leak)
动态电导越大，噪声抑制效果越强
4. Python 实现
import numpy as np
class DynamicGatedNeuron:
"""
Dynamic Gated Neuron (DGN) 实现
动态电导作为门控机制，实现自适应噪声抑制
"""
def __init__(self, tau_m=20.0, tau_g=50.0, v_th=1.0, v_reset=0.0,
g_baseline=0.1, beta=0.05, e_syn=0.0, v_rest=0.0,
r_input=1.0, dt=1.0):
self.tau_m = tau_m
self.tau_g = tau_g
self.v_th = v_th
self.v_reset = v_reset
self.g_baseline = g_baseline
self.beta = beta
self.e_syn = e_syn
self.v_rest = v_rest
self.r_input = r_input
self.dt = dt
self.v = v_rest
self.g = g_baseline
def step(self, i_syn):
"""
模拟一个时间步
Args:
i_syn: 突触输入电流
Returns:
spike: 是否发放脉冲
"""
dg = (-.g + .g_baseline) / .tau_g * .dt
dv = (-(.v - .v_rest)
- .g * (.v - .e_syn)
+ .r_input * i_syn) / .tau_m * .dt
.v += dv
.g += dg
spike =
.v >= .v_th:
spike =
.v = .v_reset
.g += .beta
spike
():
T = (input_current)
spike_train = np.zeros(T)
v_trace = np.zeros(T)
g_trace = np.zeros(T)
t (T):
i_with_noise = input_current[t] + noise_level * np.random.randn()
spike_train[t] = .step(i_with_noise)
v_trace[t] = .v
g_trace[t] = .g
spike_train, v_trace, g_trace
:
():
.tau_m = tau_m
.v_th = v_th
.v_reset = v_reset
.v_rest = v_rest
.r_input = r_input
.dt = dt
.v = v_rest
():
dv = (-(.v - .v_rest) + .r_input * i_syn) / .tau_m * .dt
.v += dv
spike =
.v >= .v_th:
spike =
.v = .v_reset
spike
():
T = (input_current)
spike_train = np.zeros(T)
v_trace = np.zeros(T)
t (T):
i_with_noise = input_current[t] + noise_level * np.random.randn()
spike_train[t] = .step(i_with_noise)
v_trace[t] = .v
spike_train, v_trace
():
T =
t = np.arange(T)
signal = + * np.sin( * np.pi * t / )
noise_levels = [, , , , , ]
results = {: [], : []}
noise noise_levels:
dgn = DynamicGatedNeuron()
spikes_dgn, _, _ = dgn.run(signal, noise)
lif = LeakyIntegrateFire()
spikes_lif, _ = lif.run(signal, noise)
rate_dgn = np.mean(spikes_dgn)
rate_lif = np.mean(spikes_lif)
noise == :
ideal_rate_dgn = rate_dgn
ideal_rate_lif = rate_lif
:
snr_dgn = ideal_rate_dgn / (ideal_rate_dgn + (rate_dgn - ideal_rate_dgn) + )
snr_lif = ideal_rate_lif / (ideal_rate_lif + (rate_lif - ideal_rate_lif) + )
results[].append(snr_dgn)
results[].append(snr_lif)
results
5. 在 SNN 网络中的应用
import torch
import torch.nn as nn
class DGNSNNLayer(nn.Module):
"""使用 DGN 神经元的 SNN 层"""
def __init__(self, n_in, n_out, tau_m=20.0, tau_g=50.0,
v_th=1.0, dt=1.0, n_steps=100):
super().__init__()
self.n_in = n_in
self.n_out = n_out
self.n_steps = n_steps
self.dt = dt
self.weight = nn.Parameter(torch.randn(n_out, n_in) * 0.1)
self.tau_m = tau_m
self.tau_g = tau_g
self.v_th = v_th
self.beta = nn.Parameter(torch.tensor(0.05))
self.g_baseline = nn.Parameter(torch.tensor(0.1))
def forward(self, x):
"""
Args:
x: 输入 [batch, n_steps, n_in] (脉冲序列)
Returns:
output: 输出 [batch, n_out] (累积脉冲计数)
"""
batch = x.size(0)
v = torch.zeros(batch, self.n_out)
g = self.g_baseline * torch.ones(batch, self.n_out)
output_spikes = torch.zeros(batch, self.n_steps, self.n_out)
for t (.n_steps):
i_syn = F.linear(x[:, t, :], .weight)
g = g + (-g + .g_baseline) / .tau_g * .dt
dv = (-(v) - g * v + i_syn) / .tau_m * .dt
v = v + dv
spikes = (v >= .v_th).()
v = v * ( - spikes)
g = g + .beta * spikes
output_spikes[:, t, :] = spikes
output_spikes.(dim=)
6. 实验验证
论文在以下基准上验证了 DGN 的优越性：
- TIDIGITS: 语音识别任务，DGN 在高噪声下显著优于 LIF
- SHD (Spiking Heidelberg Digits): 听觉时序分类
- 抗噪声任务：在加性噪声下展示增强的鲁棒性
激活关键词
- DGN
- dynamic gated neuron
- brain-inspired gating SNN
- robust spiking neural network
- dynamic conductance neuron
- noise resilient SNN
- 动态门控神经元
- 鲁棒脉冲神经网络
- 动态电导机制
Related Skills
snn-universal-approximation - SNN 万能逼近定理snn-learning-survey - SNN 学习规则综合spiking-neural-network-analysis - SNN 论文分析
