Nithyanandan Thyagarajan,Lucas Hoefs,O. Ivy Wong
Abstract:Interferometric closure invariants encode calibration-independent details of an object's morphology. Excepting simple cases, a direct backward transformation from closure invariants to morphologies is not well established. We demonstrate using simple Machine Learning models that closure invariants can aid in morphological classification and parameter estimation. We consider six phenomenologically parametrised morphologies: point-like, uniform circular disc, crescent, dual disc, crescent with elliptical accretion disc, and crescent with double jet lobes. Using logistic regression (LR), multi-layer perceptron (MLP), and random forest models on closure invariants obtained from a sparsely covered aperture, we find that all methods except LR can classify morphologies with $\gtrsim$80% accuracy, which improves with greater aperture coverage. Separately from the classification problem, given an independently confirmed class, we estimate parameters of uniform circular disc, crescent, and dual disc morphologies using simple MLP models, and parametrically reconstruct images. The estimated parameters and images correspond well with inputs, but the accuracy worsens when degeneracies between parameters are present. This independent approach to interferometric imaging under challenging observing conditions such as that faced by the Event Horizon Telescope and Very Long Baseline Interferometry in general can complement other methods in robustly constraining an object's morphology.
What problem does this paper attempt to address?
### What problem does this paper attempt to solve?
This paper aims to solve the problem of image reconstruction using interferometric closure invariants, especially in the face of sparse data and low signal - to - noise ratio conditions. Specifically, the author attempts to prove that closure invariants can be used to distinguish different image morphologies and provide an independent approach to reconstruct parameterized images through machine - learning (ML) methods.
#### Main problems
1. **Lack of direct inverse transformation**: Except for simple cases, the method of directly inverse - transforming from closure invariants to object morphologies has not been clearly established.
2. **Morphology classification and parameter estimation**: The author hopes to show that, using a simple machine - learning model, closure invariants can assist in morphology classification and parameter estimation.
3. **Application under challenging observation conditions**: Such as the observation conditions faced by the Event Horizon Telescope (EHT) and Very Long Baseline Interferometry (VLBI), these methods can supplement other methods to robustly constrain the morphology of objects.
#### Specific objectives
- **Morphology classification**: Classify images of different morphologies using closure invariants and a simple machine - learning model.
- **Parameterized image reconstruction**: Given an independently confirmed category, estimate the parameters of uniform disk, crescent - shaped and double - disk morphologies and parameterize the reconstructed image.
### Method overview
1. **Definition of closure invariants**:
- Closure phases and closure amplitudes are special interferometric quantities that are not affected by propagation and instrument effects and can eliminate some or all of the interference effects caused by the \( g_a \) term.
- Use the Abelian gauge theory formalism to obtain a complete set of independent closure invariants from co - polarization measurements.
2. **Morphology classification**:
- Six phenomenologically parameterized morphologies are considered: point - like, uniform disk, crescent - shaped, double - disk, crescent - shaped with an elliptical accretion disk and crescent - shaped with double jets.
- Use logistic regression (LR), multi - layer perceptron (MLP) and random forest (RF) models to classify closure invariants.
- The results show that, except for logistic regression, the classification accuracy of all methods reaches about 80% or more and improves as the aperture coverage increases.
3. **Parameter estimation and image reconstruction**:
- Given a known category, use a simple MLP model to estimate the parameters of uniform disk, crescent - shaped and double - disk morphologies and perform parameterized image reconstruction.
- The estimated parameters and images correspond well to the input, but the accuracy will decrease when parameter degeneracy exists.
### Summary
This paper shows the potential of closure invariants in morphology classification and parameterized image reconstruction, especially in the face of sparse data and low signal - to - noise ratio conditions, providing a new solution for interferometric imaging. This not only helps to improve the accuracy of image reconstruction, but also can reduce the influence of calibration errors and other biases in traditional methods.