Kernel Methods TRIZ Cases
- Transductive Classifier Using Maximum Entropy Discrimination
A transductive classifier trains on labeled and unlabeled data points using maximum entropy discrimination to separate categories.
- Neural Network Model Generation via Genetic Algorithm Module Optimization
A genetic algorithm optimizes neural network model generation by flexibly connecting modules and hyperparameters.
- Soft Layer Ordering in Deep Multitask Learning
Soft layer ordering applies shared neural layers in task-specific sequences to integrate information across diverse tasks.
- Anomaly Detection System with Rule Modification Interface
A machine learning anomaly detection system integrates asset-specific and non-asset-specific signals to identify potential faults in monitored assets.
- Geoscientific Data Analysis Using Order Distribution Grouping
A data analysis apparatus aligns geoscientific pair data by size and groups them using order distribution characteristics for classification.
- Medical Service Queuing System for Reducing Patient Wait Times
A queuing system directs patients to service stations based on real-time queue lengths.
- Dynamic Polling Schedule for Event Detection
A data processing method generates a polling schedule based on event probability estimates.
- Adaptive Security Platform with Polymorphic Protocol Stacks
A security platform uses machine learning classifiers to select protocol stacks and application programs for real-time threat identification.
- Text Matching Device Using Dependency Analysis for Problem Report Classification
A text matching device classifies problem reports and support information using morpheme dependency analysis.
- CMP Top Ring Vibration and Sound Analysis for Substrate Slip-Out Prediction
A machine learning model predicts substrate slip-out using top ring vibration and sound data from chemical mechanical polishing.
- Span-Level ML Model for Entity and Assertion Recognition
Joint entity recognition and assertion regression in a unified model improves detection of adverse actions while managing computational complexity.
- Adversarial Training for Machine Learning Model Data Poisoning Resistance
Adversarial training augments datasets with known attack triggers to build resilient machine learning models.
- Marker Insertion in Content Using ML Classification
A machine learning system analyzes content classification characteristics to identify optimal marker locations for precise creative placement.