菜单
Quantum Deeply-Supervised Spiking Neural Network (QDS-SNN) methodology for energy-efficient traffic sign recognition. Integrates QNNs with SNNs using TSA-LIF neurons and QACM module, achieving 99.72% accuracy with 55.77% energy reduction.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/hiyenwong/ai_collection --skill qds-snn-quantum-deeply-supervised-spiking命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | qds-snn-quantum-deeply-supervised-spiking |
| description | Quantum Deeply-Supervised Spiking Neural Network (QDS-SNN) methodology for energy-efficient traffic sign recognition. Integrates QNNs with SNNs using TSA-LIF neurons and QACM module, achieving 99.72% accuracy with 55.77% energy reduction. |
| version | 1.0.0 |
| author | Hermes Agent (Cron Job) |
| created | 2026-06-09T00:00:00.000Z |
| source | arXiv:2606.07657 |
| category | quantum-neuromorphic |
| tags | ["quantum-neural-network","spiking-neural-network","deep-supervision","energy-efficient","traffic-sign-recognition","TSA-LIF","QACM"] |
| activation_keywords | ["qds-snn","quantum spiking","deeply supervised snn","energy efficient recognition","traffic sign quantum","tsa-lif neuron"] |
QDS-SNN integrates Quantum Neural Networks (QNNs) with Spiking Neural Networks (SNNs) to overcome information loss and gradient vanishing in traditional SNN training, achieving high accuracy with significant energy efficiency improvements.
Key Performance:
# Traffic sign image preprocessing
import cv2
import numpy np
():
image = cv2.resize(image, (, ))
image = image /
image
# Temporally and Spatially Adaptive LIF implementation
class TSA_LIF:
def __init__(self, threshold, tau, alpha_temporal, alpha_spatial):
self.threshold = threshold
self.tau = tau # time constant
self.alpha_temporal = alpha_temporal # temporal adaptation
self.alpha_spatial = alpha_spatial # spatial adaptation
def forward(self, input_current, time_step, spatial_context):
# Membrane potential update
self.membrane_potential = self.tau * self.membrane_potential + input_current
# Adaptive threshold adjustment
adaptive_threshold = self.threshold + \
self.alpha_temporal * temporal_factor(time_step) + \
self.alpha_spatial * spatial_context
# Spike generation
if self.membrane_potential > adaptive_threshold:
spike = 1
self.membrane_potential = 0 # reset
else:
spike = 0
return spike
# Quantum-Assisted Classifier using PennyLane
import pennylane as qml
def create_qacm_circuit(n_qubits, n_layers):
dev = qml.device("default.qubit", wires=n_qubits)
@qml.qnode(dev)
def quantum_classifier(features):
# Encode classical features to quantum state
for i in range(n_qubits):
qml.RX(features[i], wires=i)
# Entangling layers for parallel computation
for layer in range(n_layers):
for i in range(n_qubits - 1):
qml.CNOT(wires=[i, i+1])
for i in range(n_qubits):
qml.RY(np.pi/4, wires=i)
# Measurement for classification
return [qml.expval(qml.PauliZ(i)) for i in range(n_qubits)]
return quantum_classifier
# Multi-level loss computation
def deep_supervised_loss(outputs, labels, intermediate_outputs):
total_loss = 0
# Final layer loss
final_loss = cross_entropy(outputs[-1], labels)
total_loss += final_loss
# Intermediate layer losses (deep supervision)
for i, inter_output in enumerate(intermediate_outputs):
# Weighted auxiliary loss
aux_loss = cross_entropy(inter_output, labels)
total_loss += 0.5 * aux_loss # auxiliary weight
return total_loss
| Dataset | Accuracy | Energy Consumption | Time Steps |
|---|---|---|---|
| GTSRB | 99.72% | 55.77% reduction | 6 |
| TSRD | 97.90% | 52.68% of baseline | 6 |
✅ High Accuracy: Near-perfect classification (99.72%)
✅ Energy Efficient: >50% reduction in power consumption
✅ Fast Response: Only 6 time steps for inference
✅ Gradient Stability: Deep supervision prevents vanishing gradients
✅ Quantum Parallelism: Expressive representations without overhead
⚠️ Hardware Dependency: Requires quantum simulation or actual quantum hardware
⚠️ Training Complexity: Multi-level supervision increases training time
⚠️ Dataset Specific: Optimized for traffic signs; may need adaptation for other tasks
Activation: Use when developing energy-efficient vision systems, quantum-neuromorphic applications, or solving SNN gradient problems. Keywords: quantum spiking, TSA-LIF, QACM, traffic sign recognition, deeply supervised SNN.