Biological Models TRIZ Cases
- Power System Fault Point Locater Using Existing Relay Data
A fault point locater system uses existing substation relay data to quantify fault probability without extra sensors.
- Deferred Neural Lighting for Augmented Image Generation
Neural deferred rendering model processes panoramic source lighting and augmented image buffers to generate photorealistic composite images.
- Style Transfer for Object Detection Robustness
Generates diverse style representations to train object detection models, reducing annotation costs.
- Generative UI Generation via LLM Query Decomposition
A generative user interface system maps queries to variables using large language models to produce dynamic code.
- Composite Loss Parameter for Imbalanced Anomaly Detection
A composite loss parameter merges anomaly detection and classification prediction models, resolving accuracy drops in imbalanced datasets.
- Endpoint Prediction in Speech Processing Systems
Preliminary endpoint prediction by the ASR component allows concurrent transcript publishing, resolving latency versus accuracy trade-offs.
- Deep Learning Road Extraction via Multi-Task Autoencoder
Multi-task autoencoders segment roads and detect objects simultaneously, resolving the trade-off between extraction accuracy and processing complexity.
- Federated Learning Neural Graphical Models for Privacy
Neural graphical models aggregate client distributions into a global dependency graph without sharing raw data.
- Privacy-Preserving Data Transformation in Machine Learning Models
A machine learning data transformation module suppresses sensitive attributes while preserving useful information through specialized neural networks.
- Self-supervised Velocity Learning for Autonomous Systems
A machine learning model estimates object velocities from measured RADAR data to refine detection parameters.
- SCouT Transformer for Synthetic Counterfactuals
SCouT uses causal maps to predict intervention effects despite noisy longitudinal data.
- Federated Learning Model Training with Encrypted Data Partitioning
A modeling component trains inferential models using horizontally and vertically partitioned data via random decision trees.
- Distributed Training Node Selection via Waiting and Execution Time Scoring
Information processing program calculates scores from waiting and execution times to determine optimal node counts for distributed training.
- NPC Behavior Control Using Reinforcement Learning for RPG Skill Casting
A reinforcement learning network extracts game state features to generate action probability distributions for non-player character skill casting.
- Nested ML Models for Exam Fraud Risk Assessment
Nested machine learning models generate aggregate risk scores to detect fraud in exam delivery systems.
- Graph-Based Synthetic Training Data Generation for Autonomous Driving
Transform sensor data into graph structures to generate diverse synthetic training examples for machine learning models.
- Neural Network Artificial Defect Data Generation for Vision Inspection
A neural network model generates artificial defect data by transforming input sub-data to expand training sets.
- Chatbot with LLM and Vectorized Database for Card Benefit Identification
A chatbot uses a large language model and vectorized database to identify payment card benefits through natural language interaction.
- Domain-Specific Generative ML Models for IT Incident Resolution
A domain-specific generative machine learning model refines incident data to generate accurate resolutions.
- AI CBRN Threat Prediction via Sensor Data Correction
U-Net technology refines diffusion modeling to deliver reliable CBRN threat predictions despite complex urban sensor constraints.
- Modular Base Station Orchestration for Demand Spikes
Dynamic base station positioning manages network demand spikes without increasing manual system complexity.
- AI Code Translation Visualizing Low Confidence Regions
Color-coded visual cues correlate AI confidence with lint violations to pinpoint translation errors and reduce manual review time.
- Radar Object Detection Using Neural Network Signal Processing
Neural networks replace complex signal processing pipelines with learned parameters, reducing computational load while maintaining high detection accuracy.
- Residual Neural Network for Video Temporal Redundancy
A residual neural network generates frame representations by combining clip and residual data.
- Machine Learning Model for Data Credibility Assessment
A trained machine learning model assesses data credibility by integrating property information with recipient-specific reliability lists.
- LLM Process Control Analytics Automation
AI system generates analytics instructions from natural language queries to create customized visualizations.
- Noise-Level Segmentation for Machine Learning Model Training
Segmenting training data by noise level enables tailored model development without discarding valuable information.
- Anomaly Detection in Text Using Variational Autoencoders
Variational autoencoders analyze visual text features to detect anomalies, reducing false positives from font variations and overfitting.
- Personalized Facial Expression Classification via Synthesized AU Datasets
Synthesizing personalized facial images with specific Action Unit combinations to train machine learning models for expression recognition.
- Dynamic Survival Analysis Model with Soft Labels
A prediction model constructs soft labels from estimated probability distributions to improve sample efficiency.
- Neural Network Voice Type Detection Using LPC Features
A neural network classifies vocal efforts into distinct voice types using linear predictive coding coefficients as input features.
- Vehicle Sensor Neural Network Interference Detection
Monitoring neural network analysis behavior detects sensor interferences from environmental factors to maintain automation reliability.
- Blendshape Weight Prediction Using Rendered Avatar Training Data
Rendered avatar training images eliminate manual labeling bottlenecks, enabling accurate blendshape weight prediction for HMD facial avatars.
- Neural Network Routing in Network Devices
Trained neural network determines packet forwarding recommendations using programmable hardware to resolve static routing inefficiencies.
- Real-Time Avatar Animation Audio Separation
Deep neural network separates vocal and non-vocal audio components to drive synchronized avatar animations.