Examples#
Here we show some real world examples that use PyTorch Network Loader.
See the Jupyter Notebook for a short demo.
Variational Autoencoder with Normalizing Flow for X-Ray Spectral Fitting#
They introduce a probabilistic model, utilizing a variational autoencoder with a normalizing flow, trained to adopt a physical latent space. This neural network produces predictions for spectral-model parameters as well as their full probability distributions. Their implementations result in a significant improvement in spectral reconstructions over a previous deterministic model while performing three orders of magnitude faster than traditional methods.
Measuring the Dark Matter Self-Interaction Cross-Section with Deep Compact Clustering for Robust Machine Learning Inference#
This project implements a clustering neural network for constraining the dark matter self-interaction cross-section from galaxy clusters using cosmological simulations. The method builds an interpretable latent space where galaxy clusters are grouped by physical similarity, enabling robust parameter estimation in sparsely sampled parameter spaces while providing confidence measurements to detect out-of-domain data and exploration of secondary features in the data. They apply this method to the BAHAMAS-SIDM and DARKSKIES simulations and demonstrate accurate recovery of the self-interaction cross-section when the test data lies within the training domain and reliable rejection of foreign datasets, offering a blueprint for robust machine learning for scientific inference.
Rapid Spectral Parameter Prediction for Black Hole X-Ray Binaries using Physicalized Autoencoders#
This project presents a new semi-supervised autoencoder neural network for parameter prediction and spectral reconstruction of black hole X-ray binaries (BHBs), showing an improvement of up to a factor of 2,700 in speed, while maintaining comparable accuracy. The approach maps the spectral features from the numerous outbursts catalogued by NICER and generalises them to new systems for efficient and accurate spectral fitting. The effectiveness of this approach is demonstrated in the spectral fitting of BHBs and holds promise for use in other areas of astronomy and physics for categorising large dataset.
Simulation-Based Inference for Exoplanet Atmospheric Retrieval: Insights from Winning the Ariel Data Challenge 2023 Using Normalizing Flows#
This project presents novel machine learning models developed by the AstroAI team for the Ariel Data Challenge 2023, where one of the models secured the top position among 293 competitors. Leveraging Normalizing Flows, their models predict the posterior probability distribution of atmospheric parameters under different atmospheric assumptions. Moreover, they introduce an alternative model that exhibits higher performance potential than the winning model, despite scoring lower in the challenge. These findings highlight the need to reevaluate the evaluation metric and prompt further exploration of more efficient and accurate approaches for exoplanet atmosphere spectra analysis. Finally, they present recommendations to enhance the challenge and models, providing valuable insights for future applications on real observational data. These advancements pave the way for more effective and timely analysis of exoplanet atmospheric properties, advancing our understanding of these distant worlds.