Machine Learning TRIZ Cases
- Genetic Algorithm Rule Labeling for Persistent State Memory
A genetic algorithm incorporates a label field in rule outputs to maintain the reason for previous assertions.
- ML Workload Forecasting for Cloud Worker Instance Scaling
Machine learning model predicts time series workload to compute required worker instances, reducing allocation request throttling and timeout exceptions.
- Network Switch Compute Subsystem for Distributed AI Gradient Aggregation
Embedding compute logic in network switches performs collective reduction operations on gradients, halving communication latency and overhead.
- Network Scanning System for Software Copyright Detection
A network scanning system uses a trained similarity algorithm to identify source code matches across the internet.
- Data Processing Apparatus for Abnormality Cause Analysis
A data processing apparatus classifies samples using focus thresholds to identify abnormality causes.
- Proxy Prediction Models for On-Device Machine Learning
Server-generated proxy prediction models enable client devices to select personalized machine learning parameters without sharing private user data.
- Wearable Sensor Workforce Risk Assessment System
A wearable sensor system derives individual risk scores to quantify workplace safety metrics.
- Intelligent Capacity Planning for High Load Variance
ML subsystem predicts capacity to intercept peak requests, preventing chokepoints during high load variance.
- Secure Check Processing System with Hidden Image Verification
A secure check processing system embeds unique hidden images and watermarks on stored check images to verify origin and prevent unauthorized alterations.
- Adaptive GUI for Motion Contexts Using ML Prediction
Machine learning models predict user responses from motion data to generate adaptive graphical user interface variations.
- Fixed-Point Neural Network Quantization via Statistical Fraction Length
A method generates fixed-point neural networks by deriving statistical characteristics from post-activation values to determine optimal fraction lengths.
- Transformer Row Embeddings for Tabular Data Semantic Clustering
A transformer model generates mathematical vector embeddings for data table rows to capture semantic meaning across mixed data types.
- Cross-Node AI Session Handover in RAN
Preserve cross-node machine learning session continuity during user equipment handovers by relaying context between network entities.
- IoT Indirect Interaction System for Smart Home Task Coordination
An IoT system enables indirect user interactions through intelligent context prediction and automated interface routing.
- Cloud Network Fault Management via Performance-Based Thresholds
Testing tools measure target server performance to set accurate thresholds, resolving the contradiction between alarm reliability and measurement precision.
- Local ANN Dataset Memory Management for Efficient Processing
Local storage of neural network datasets reduces data movement and power consumption while improving throughput capacity.