Biological Models TRIZ Cases
- ResUnet Network for Mel Spectrum Feature Enhancement
A ResUnet network model processes Mel spectrum features through down sampling and up sampling to restore lost information.
- Certifiable Piloting Rules via Reinforcement Learning Segmentation
Segmenting reinforcement learning from formal verification produces certifiable piloting rules that resolve inconsistent command generation.
- Bidirectional RNN for Low-Frequency Abnormal Traffic Detection
Bidirectional recurrent neural networks predict next traffic packets to detect low-frequency anomalies without stopping monitoring.
- Automated Root Cause Analysis for Telecommunications Networks
Machine learning algorithms analyze incomplete performance monitoring data to identify root causes of signal degradation without requiring expert intervention.
- AI Query Context Evaluation for Reducing Bandwidth
An AI system evaluates item attributes across multiple user queries to select relevant digital components.
- Memristor Crossbar On-Chip Training via Analog Error Propagation
Analog neuromorphic circuit uses resistive memories to propagate input voltage signals for parallel computation.
- Training Dynamics Prediction Model for Active Learning
Pre-calculating training dynamics reduces computational cost while maintaining inference accuracy for active machine learning data selection.