Dimensions overall score 7.0

GitHub Code Pulse

No public code linked for this paper yet.

• LiteInception reduces input sensor channels from 23 to 15 using a multi-method fusion strategy based on mutual information, gradient analysis, and SE attention weights.method 90%
• The 1+1 branch LiteInception architecture compresses InceptionTime parameters by 70% and accelerates CPU inference by over 8x with less than 3% F1 loss.result 90%
• LiteInception achieves 81.92% fault detection accuracy with 83.24% recall on the NGAFID dataset.result 90%
• The framework achieves 77.00% fault identification accuracy on the NGAFID dataset.result 90%
• Knowledge distillation enables the same lightweight model to adapt to different scenarios by switching training strategies.method 85%
• The framework uses a two-stage cascaded architecture with Stage 1 performing high-recall fault detection and Stage 2 conducting fine-grained fault classification.method 90%
• A dual-layer interpretability framework integrating four attribution methods provides traceable evidence chains of 'which sensor x which time period.'method 85%
• The framework decouples optimization objectives and enables on-demand allocation of computational resources through its two-stage architecture.technical 80%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card