User Guide

• 1. Getting started on PC (Linux/MacOS/Windows)
  • 1.1. Prerequisites
  • 1.2. Install scikit-learn
  • 1.3. Install emlearn
  • 1.4. Create model in Python
  • 1.5. Use in C code
  • 1.6. Try it out
  • 1.7. Next
• 2. Getting started on hardware (Arduino)
  • 2.1. Prerequisites
  • 2.2. Create Arduino project
  • 2.3. Create model in Python
  • 2.4. Copy model files into Arduino
  • 2.5. Install emlearn as Arduino library
  • 2.6. Build and upload the code
  • 2.7. Test it out
  • 2.8. Next
• 3. Getting started on Zephyr RTOS
  • 3.1. Prerequisites
  • 3.2. Add emlearn to your Zephyr project
  • 3.3. Enable emlearn in your Zephyr project
  • 3.4. Create model in Python
  • 3.5. Example application code
  • 3.6. Build and run in simulator
• 4. Getting started on browser (WASM/Emscripten)
  • 4.1. Prerequisites
  • 4.2. Install scikit-learn
  • 4.3. Install emlearn
  • 4.4. Create model in Python
  • 4.5. Use in C code for WASM module
  • 4.6. Load WASM module in webpage
  • 4.7. Try it out
  • 4.8. Next
• 5. Getting started with MicroPython
• 6. Platform support
  • 6.1. Software frameworks and RTOS
  • 6.2. Microcontroller hardware platforms
  • 6.3. Embedded OS platforms
  • 6.4. Other programming languages
  • 6.5. More
• 7. Feature extraction
  • 7.1. Time-domain features
  • 7.2. Statistical summaries
  • 7.3. Digital filters
  • 7.4. Spectrum (frequency domain)
  • 7.5. Spectrogram (time-frequency domain)
  • 7.6. Integrating feature extraction
• 8. Classification
  • 8.1. Applications
  • 8.2. Classification models
  • 8.3. Related
• 9. Regression
  • 9.1. Applications
  • 9.2. Regression models
  • 9.3. Related
• 10. Anomaly Detection
  • 10.1. Applications
  • 10.2. Trade-off between False Alarms and Missed Detections
  • 10.3. Selecting anomaly threshold
  • 10.4. Anomaly Detection models
  • 10.5. Outlier detection for handling unknown data
• 11. Event Detection
  • 11.1. Applications
  • 11.2. Binary or multi-class
  • 11.3. Sliding time windows
  • 11.4. Classification models for Event Detection
  • 11.5. Training with class imbalance
  • 11.6. Trade-off between False Alarms and Missed Detections
  • 11.7. Evaluation metrics
  • 11.8. Frame-wise evaluation
  • 11.9. Event-wise evaluation
• 12. Model optimization
  • 12.1. Goals for efficient models
  • 12.2. Measuring model costs
• 13. Tree-based models
  • 13.1. Basic usage
  • 13.2. Probability output
  • 13.3. Regression
  • 13.4. Inference method: Inline vs loadable
  • 13.5. Feature representation
  • 13.6. Hard voting (majority) vs soft voting (proportions)
  • 13.7. Target quantization and leaf-deduplication
  • 13.8. Optimizing model complexity
  • 13.9. Optimization of features