8. Advanced Features
Tiny ML Tensorlab includes several advanced features to help you build more accurate and efficient models. This section covers these capabilities in detail.
Contents
- 8.1. Neural Architecture Search
- 8.1.1. Overview
- 8.1.2. When to Use NAS
- 8.1.3. Enabling NAS
- 8.1.4. Configuration Options
- 8.1.5. Search Types
- 8.1.6. Search Space
- 8.1.7. Running NAS
- 8.1.8. Output Files
- 8.1.9. Interpreting Results
- 8.1.10. Using NAS Results
- 8.1.11. Advanced Configuration
- 8.1.12. Example: Finding Optimal Arc Fault Model
- 8.1.13. Computational Cost
- 8.1.14. Best Practices
- 8.1.15. Next Steps
- 8.2. Quantization
- 8.2.1. Overview
- 8.2.2. Configuration Parameters
- 8.2.3. Quantization Modes
- 8.2.4. Quantization Methods
- 8.2.5. Bit Widths
- 8.2.6. NPU Quantization Requirements
- 8.2.7. Output Files
- 8.2.8. Accuracy Comparison
- 8.2.9. Troubleshooting Accuracy Loss
- 8.2.10. Best Practices
- 8.2.11. Example: Full Quantization Workflow
- 8.2.12. Memory Savings
- 8.2.13. Performance Impact
- 8.2.14. Next Steps
- 8.3. Feature Extraction
- 8.3.1. Overview
- 8.3.2. Feature Extraction Pipeline
- 8.3.3. Configuration Parameters
- 8.3.4. Preset System
- 8.3.5. Available Presets
- 8.3.6. Data Processing Transforms
- 8.3.7. Feature Extraction Transforms
- 8.3.8. Custom Feature Extraction
- 8.3.9. Multi-Channel Data
- 8.3.10. Forecasting Configuration
- 8.3.11. Data Augmentation
- 8.3.12. Choosing the Right Preset
- 8.3.13. Performance Impact
- 8.3.14. On-Device Feature Extraction
- 8.3.15. Example Configurations
- 8.3.16. Best Practices
- 8.3.17. Next Steps
- 8.4. Goodness of Fit
- 8.4.1. Overview
- 8.4.2. Enabling GoF Test
- 8.4.3. Running the Test
- 8.4.4. Output Files
- 8.4.5. Understanding the Visualizations
- 8.4.6. Interpreting Results
- 8.4.7. 8-Plot Analysis
- 8.4.8. Common Patterns
- 8.4.9. Actionable Insights
- 8.4.10. Example: Motor Fault GoF Analysis
- 8.4.11. GoF Without Training
- 8.4.12. Comparing Feature Extraction
- 8.4.13. Best Practices
- 8.4.14. Limitations
- 8.4.15. Next Steps
- 8.5. Post-Training Analysis
- 8.5.1. Overview
- 8.5.2. Enabling Analysis
- 8.5.3. Output Files
- 8.5.4. Confusion Matrix
- 8.5.5. ROC Curves
- 8.5.6. Class Score Histograms
- 8.5.7. FPR/TPR Thresholds
- 8.5.8. Classification Report
- 8.5.9. Error Analysis
- 8.5.10. Quantized vs Float Comparison
- 8.5.11. Regression Analysis
- 8.5.12. Anomaly Detection Analysis
- 8.5.13. Custom Analysis Scripts
- 8.5.14. Generating Reports
- 8.5.15. Example: Complete Analysis Configuration
- 8.5.16. Best Practices
- 8.5.17. Troubleshooting Low Accuracy
- 8.5.18. Next Steps
8.6. Feature Overview
Neural Architecture Search (NAS)
Automatically discover optimal neural network architectures for your dataset. NAS can optimize for memory usage or computational efficiency.
Preset sizes:
s,m,l,xl,xxlOptimization modes: Memory or Compute
GPU recommended for practical use
Quantization
Reduce model size and improve inference speed through quantization:
QAT (Quantization-Aware Training) - Best accuracy
PTQ (Post-Training Quantization) - Faster, no retraining
Weight bit-widths: 2-bit, 4-bit, 8-bit
Feature Extraction
Transform raw time-series data into meaningful features:
FFT (Fast Fourier Transform)
Binning and normalization
Haar and Hadamard wavelets
Logarithmic scaling
Goodness of Fit Test
Evaluate whether your dataset is suitable for classification before training. Uses PCA and t-SNE visualization to assess class separability.
Post-Training Analysis
Understand model performance with:
ROC curves for classification
Confusion matrices
FPR/TPR threshold analysis
PCA visualization of feature-extracted data