Genetic Models TRIZ Cases
- ML Architecture Search Framework for Hardware-Aware Multi-Metric Optimization
Automatically searches ML architectures across hardware constraints and performance metrics to cut design time and compute use.
- Multi-Task Neural Network Training With Frozen Shared Blocks
Frozen shared blocks and task-specific updates let one neural network learn multiple tasks while limiting forgetting, overhead, and latency.
- FSM Stateflow Optimization for Correct and Usable Behavior
Iterative scoring with domain rules, genetic search, and local search improves FSM stateflow correctness, usability, and expected behavior.
- Oligonucleotide Data Encoding to Disrupt Repeating Sequence Patterns
Transverse encoding and encryption disrupt repeating and self-folding nucleotide patterns, reducing sequencing errors in retrieved data.
- Neural Network Task Segmentation to Prevent Catastrophic Forgetting
Segmented task networks limit catastrophic forgetting and training interference.
- Machine Learning Architecture Search With Proxy-Based Evaluation
Evolutionary search uses proxy training and final validation to optimize ML architectures across hardware platforms.
- Self-Evolving AI Playstyle Models from Gameplay and Emotion Data
Gameplay decisions and emotional states drive self-updating playstyle models for more tailored, evolving player experiences.
- Pyramid Layer Architecture Search for Accuracy-Latency Tradeoffs
Automated pyramid-layer architecture search balances multi-scale detection accuracy with computational latency for anytime inference.
- Universal Information Flow Tracking Across Heterogeneous Networks
A method tracks information flows through multiple network systems by creating standardized descriptor sets for interacting nodes.
- Hierarchical Model Tree for Real-Time Object Recognition
A hierarchical model tree groups extracted image tokens to identify objects in target images.
- Program Generation Apparatus Constraint Validation Overfitting
A program generation device creates software satisfying input-output pairs and validates outputs against random inputs.
- Phylogenetic Generative Models for Biological Association Analysis
Generative models account for phylogenetic structure to identify variable associations, resolving confounding effects from evolutionary relationships.
- Combinatorial Solver Benchmarking via Randomized Input Sequences
Randomized input sequences enable unbiased evaluation of combinatorial solver performance across diverse problem instances.
- Phase Space Prediction for Chaotic Systems
Predict chaotic system states by embedding time series data into a reconstructed phase space and analyzing the maximum Lyapunov exponent.
- Disk Drive Profile Table Tuning for Seek Time Optimization
Constant offset values in a programmable profile table adjust seek time performance, reducing manufacturing workload and costs across diverse drive types.
- Offloading Server Automates GPU Code Generation
An offloading server analyzes application code to generate optimized accelerator directives for parallel processing.
- Offload Server Automatic FPGA Processing via High-Level Synthesis
An offload server uses high-level synthesis to automatically generate parallel processing patterns for programmable logic devices.
- Social Matching via Behavioral Monitoring and Mutual Contact Verification
A matching system monitors user activities to suggest matches based on observed behaviors rather than self-reported profiles.
- Genetic Convolutional Neural Network Layer for Bias-Free Tool Generation
A genetic convolutional neural network layer automatically generates and weights processing tools using evolutionary algorithms.
- Clinical Decision Support Rule Database Automation
Ontology-based inference and self-evolutionary algorithms automate clinical rule database updates, eliminating expert intervention overheads.
- Genetic Algorithm for Database Access Structure Selection Under Storage Constraints
Optimal-storage profiling and elitism guide the search to minimize cost, balancing query speed against limited storage space.
- Morphological Genome for Universal Form Generation
A morphological genome encodes form variations through independent genes mapped in higher-dimensional Euclidean space.
- Coordinated Reliability Management for Virtual Machine Migration
A coordinated reliability management system migrates virtual machines to hosts with appropriate reliability levels.