Self-supervised Learning with Physics-aware Neural Networks I: Galaxy Model Fitting

Abstract

Estimating the parameters of a model describing a set of observations using a neural network is in general solved in a supervised way. In cases when we do not have access to the model's true parameters this approach can not be applied. Standard unsupervised learning techniques on the other hand, do not produce meaningful or semantic representations that can be associated to the model's parameters. Here we introduce a self-supervised hybrid network that combines traditional neural network elements with analytic or numerical models which represent a physical process to be learned by the system. Self-supervised learning is achieved by generating an internal representation equivalent to the parameters of the physical model. This semantic representation is used to evaluate the model and compare it to the input data during training. The Semantic Autoencoder architecture described here shares the robustness of neural networks while including an explicit model of the data, learns in an unsupervised way and estimates, by construction, parameters with direct physical interpretation. As an illustrative application we perform unsupervised learning for 2D model fitting of exponential light profiles.

References in corpus (9)

• ADADELTA: An Adaptive Learning Rate Method
• Rotation-invariant convolutional neural networks for galaxy morphology prediction
• Discovering physical concepts with neural networks
• Improving galaxy morphologies for SDSS with Deep Learning
• Star-galaxy Classification Using Deep Convolutional Neural Networks
• GALAPAGOS: From Pixels to Parameters
• Deep learning for galaxy surface brightness profile fitting
• Classifying the Large Scale Structure of the Universe with Deep Neural Networks
• Galaxy shape measurement with convolutional neural networks

Cited by in corpus (6)

• The DAWES review 10: The impact of deep learning for the analysis of galaxy surveys
• Solving inverse-PDE problems with physics-aware neural networks
• A black box for dark sector physics: Predicting dark matter annihilation feedback with conditional GANs
• Solving the Pulsar Equation using Physics-Informed Neural Networks
• Predicting Change, Not States: An Alternate Framework for Neural PDE Surrogates
• A self-supervised, physics-aware, Bayesian neural network architecture for modelling galaxy emission-line kinematics