Shrink Calibration Plugin for Pioreactor

This plugin provides functionality to shrink multiple calibrations across workers in a Pioreactor cluster. It uses a regularized optimization process to fit a unified polynomial model while adjusting for worker-specific variations.

Features

• Collects calibration data from multiple workers in a Pioreactor cluster.
• Fits a regularized polynomial model.
• Allows user interaction to select calibrations and customize parameters.
• Automatically distributes the resulting calibration back to the workers.

Prerequisites

Before using this plugin, ensure you have:

• A Pioreactor cluster with workers set up and active.
• Calibrations already recorded on each worker for the specified device.
• The Pioreactor software stack installed on your system.

Installation

There are three options:

1. Install this plugin on your leader via the UI's Plugin page (requires internet access)
2. Install this plugin on your leader via pio plugins install pioreactor-calibration-shrinkage (requires internet access)
3. Or copy the __init__.py file's contents into a new .py file in the ~/.pioreactor/plugins directory on your leader.

Usage

Running the Plugin

Use the following command to start the shrink calibration process:

pio calibrations shrinkage --device <DEVICE>

Replace <DEVICE> with the device you want to shrink calibrations for (e.g., pump, sensor).

Workflow

1. Calibration Selection: The plugin will list available calibrations on each worker. Select the calibrations you wish to include in the shrinkage process.
2. Parameter Configuration:
   • Polynomial Degree: Specify the degree of the polynomial for the unified model.
   • Closeness Parameter: Adjust the regularization parameter (lambda_a) to control how closely individual calibrations align with the average.
3. Model Fitting: The plugin fits the model and displays the results. If you're not satisfied, you can adjust parameters and refit.
4. Calibration Distribution: Once satisfied, the plugin distributes the new calibration data back to the workers.

Example

pioreactor calibrations shrinkage --device media_pump

• Follow the prompts to select calibrations and configure the fitting parameters.
• The plugin will output logs detailing the progress and results.

How It Works

1. Data Collection: The plugin retrieves calibration data from each worker using the Pioreactor API.
2. Model Fitting: It fits a polynomial model using regularized optimization, adjusting worker-specific scalars (A) and the polynomial coefficients (w).
3. Calibration Creation: A new calibration is created for each worker, incorporating the unified model and worker-specific adjustments.
4. Calibration Distribution: The new calibration is posted back to each worker for use.

Support

If you encounter any issues or have questions, please visit the Pioreactor community forums or contact support.

Happy calibrating! 🎉