Skip to content

ML4SCI/DeepLense

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ML4Sci x DeepLense

1. Background

We at DeepLense explore cutting-edge Machine Learning techniques for the study of Strong Gravitational Lensing and Dark Matter Sub-structure. We use both simulated and real lensing images, for a variety of tasks, using a variety of techniques.

We also actively mentor Google Summer of Code (GSoC) projects, that you can find listed here.

  1. Find below a description of gravitational lensing and dark matter sub-structure.
  2. Section 2 contains a detailed description of the datasets used in the various projects
  3. Section 3 beins with a short description followed by an expansion on the various (GSoC) projects conducted at DeepLense

1.1 Gravitational Lensing

Gravitational lensing is the phenomenon of the bending of light in the gravity of a massive celestial object (such as a massive galaxy or a group of galaxies); the object essentially behaving as a cosmic lens. We, as a result see the distorted image(s) of light sources (typically another galaxy) behind it.

Lensing Schematic

The dynamics of lensing depends on both the composition of the lens and the nature of the source. We explore different lens models and source light profiles including real galaxy images in DeepLense, that you can find here.

1.2 Dark Matter and Sub-structure

In DeepLense, we’re mainly dealing with three kinds of simulated Dark Matter:

  • Axion Dark Matter (Vortex): Axions are hypothetical particles that are considered as candidates for dark matter. In the context of axion dark matter, vortex substructures refer to specific topological features that can form in the distribution of axion fields.
  • Cold Dark Matter (Subhalo): This model suggests that dark matter consists of slow-moving particles. In the CDM paradigm, smaller clusters of dark matter, known as subhalos, are approximated as “point masses.” This simplification facilitates computational modeling by treating these subhalos as singular points in the overall distribution of dark matter.
  • No-Substructure Dark Matter: Unlike the CDM model, the “no-substructure” approach assumes that dark matter is evenly spread out, devoid of any smaller-scale clusters or sub-halos. This stands in stark contrast to the hierarchical structuring and layering of sub-halos within larger halos as predicted by CDM models.

2. Datasets

All datasets are constructed using Lenstronomy, by Michael W. Toomey, as presented in their repository.

Dataset Lens model Light Profile Modelling strategy
Model 1 dataset Sheared Isothermal Elliptical lens Sérsic light profile Gaussian point spread function, Gaussian and Poissonian noise for SNR ~ 25, Axion DM and CDM substructure appended to base halo to create 3 sub-structure classes
Model 2 dataset Sheared Isothermal Elliptical lens Sérsic light profile Euclid's observation characteristics, Axion DM and CDM substructure appended to base halo to create 3 sub-structure classes
Model 3 dataset Sheared Isothermal Elliptical lens Sérsic light profile HST's observation characteristics, Axion DM and CDM substructure appended to base halo to create 3 sub-structure classes
Model 4 dataset Two Isothermal Elliptical lenses Three-channel real galaxy images Euclid's observation characteristics, Axion DM and CDM substructure appended to base halo to create 3 sub-structure classes

3. Projects

Project compositions Techniques compositions

Project Name Contributor Task ML Techniques Repository Link Blog Post
DeepLense Classification Using Vision Transformers Archil Srivastava Sub-structure classification Vision Transformer Variants Click Here Click Here
Unsupervised Domain Adaptation Mriganka Nath Sub-structure classification Unsupervised Domain Adaptation Click Here Click Here
Lensiformer: A Relativistic Physics-Informed Vision Transformer Architecture for Dark Matter Morphology in Gravitational Lensing Lucas Jose Sub-structure classification Physics-Informed Transformer Click Here Click Here
Physics Informed Neural Network for Dark Matter Morphology Ashutosh Ojha Sub-structure classification Physics-Informed Transformer Click Here Click here
Learning Representation Through Self-Supervised Learning on Real Gravitational Lensing Images Sreehari Iyer Sub-structure classification Self-supervised Learning Click Here Click Here
Contrastive Learning vs BYOL Yashwardhan Deshmukh Sub-structure classification Self-supervised Learning Click Here Click Here
Domain Adaptation for Simulation-Based Dark Matter Searches Using Strong Gravitational Lensing Marcos Tidball Sub-structure classification Unsupervised Domain Adaptation Click Here Click Here
Equivariant Neural Networks for Dark Matter Morphology with Strong Gravitational Lensing Apoorva Singh Sub-structure classification Equivariant Neural Networks Click Here Click Here
Equivariant Neural Networks for Dark Matter Morphology with Strong Gravitational Lensing Geo Jolly Sub-structure classification Equivariant Neural Networks Click Here
Self-Supervised & Supervised Learning Kartik Sachdev Sub-structure classification Self-Supervised Learning Click Here Click Here
Updating the DeepLense Pipeline Saranga K Mahanta Sub-structure classification, Anomaly Detection and Regression Autoencoder Variants Click Here Click here
DeepLense Regression Yurii Halychanskyi Property estimation through regression Residual Networks, Transformer Variants Click Here First and second projects
DeepLense Regression Zhongchao Guan Property estimation through regression Residual Networks, Transformer Variants Click Here Click Here
Single Image Super-Resolution with Diffusion Models Atal Gupta Super-Resolution Diffusion Models Click Here Click Here
Super-Resolution for Strong Gravitational Lensing Pranath Reddy Super-Resolution Residual Networks, Conditional Diffusion Models Click Here Click Here
Physics-Informed Unsupervised Super-Resolution of Strong Lensing Images Anirudh Shankar Super-Resolution Physics-Informed SISRs Click Here Click Here

3.1 Dark matter sub-structure classification

Classification of lensing images into their intrinsic dark matter sub-structure can reveal information about their characteristics and ways to study them. DeepLense presents an extensive array of techniques to study dark matter substructure involving both the simulated and real galaxy datasets.

3.1.1 DeepLense Classification Using Vision Transformers

Archil Srivastava, as part of their GSoC 2022 project, explores the potency of variants of Vision Transformers (EfficientNet, ViT, ConViT, CrossViT, Bottleneck Transformers, EfficientFormer, CoaT, CoAtNet and Swin) on the classification of dark matter substructure on the three datasets, Model 1, 2 and 3.

3.1.2 Unsupervised Domain Adaptation

Mriganka Nath employs unsupervised domain adaptation techniques as part of their GSoC 2022 project such as ADDA, AdaMatch and self-ensembling from simulated lensing images to classify dark matter substructure of real lensing images from the Hyper Suprime-Cam (HSC) Subaru Strategic Program Public Data Release 3.

3.1.3 Lensiformer: A Relativistic Physics-Informed Vision Transformer Architecture for Dark Matter Morphology in Gravitational Lensing

Lucas Jose bridges the gap between relativistic Physics principles and machine learning models in their GSoC 2023 project, where they develop a Physics-Informed transformer architecture that outperforms its traditional counterparts.

3.1.4 Physics Informed Neural Network for Dark Matter Morphology

Ashutosh Ojha builds on to improve Lucas' Physics-Informed transformer in their GSoC 2024 project, through the inclusion of Physics-Informed pre-processing, and a GradCAM visualization allowing for the interpretation of its working.

3.1.5 Learning Representation Through Self-Supervised Learning on Real Gravitational Lensing Images

Sreehari Iyer evaluates Transformer-based self-supervised learning techniques through their GSoC 2024 project, utilizing the real-world strong gravitational lensing dataset, Model 4. ViT-S and ViT-B are selected as the backbone and SimSiam, DINO and iBOT for self supervised training.

3.1.6 Contrastive Learning vs BYOL

Yashwardhan Deshmukh compares the performance of the self-supervised learning techniques in their GSoC 2023 project, Contranstive Learning and Bootstrap Your Own Latent (BYOL) on the three datasets, Model 1, 2 and 3.

3.1.7 Domain Adaptation for Simulation-Based Dark Matter Searches Using Strong Gravitational Lensing

Marcos Tidball performs unsupervised domain adaptation in their GSoC 2021 project, to mitigate the poor generalization of models trained on simulated data to real lensing images using ADDA, Self-Ensemble, CGDM and AdaMatch. Their work has been published as a paper in the Astrophysical Journal.

3.1.8 Equivariant Neural Networks for Dark Matter Morphology with Strong Gravitational Lensing

Apoorva Singh, in their GSoC 2021 project and Geo Jolly, in their GSoC 2023 project exploit the inherent symmetries present in the strong lensing system (such as rotations and reflections) using equivariant neural networks, to extract dark matter sub-structural information.

3.1.9 Classification : Self-Supervised & Supervised Learning

Kartik Sachdev, through their GSoC 2022 and GSoC 2023 projects performs benchmarking of an extensive variety of vision transformer architectures, and contrasting of ten supervised and two self-supervised learning frameworks on the classification of dark matter substructure. They use the three datasets, Model 1, 2 and 3.

3.1.10 Updating the DeepLense Pipeline

Saranga K Mahanta, in their GSoC 2022 project conducts a study on strong lensing through a variety of tasks on strong lensing images including classification, anomaly detection and regression using several different neural network architectures, on the three datasets, Model 1, 2 and 3.

3.2 Dark matter property estimation through regression

Another means of dark matter study through strong lensing is through the approximation of their properties. Yurii Halychanskyi and Zhongchao Guan approximate the mass density of vortex substructure of dark matter condensates on the three datasets, Model 1, 2 and 3. Yurii uses the ResNet18Hybrid and CmtTi architectures in their GSoc 2021 and 2022 projects, while Zhongchao demonstres with ResNet18, ViT, CNN-T, MobileNet V2 and CvT-13, in their GSoc 2022 project.

3.3 Super-resolution of lensing images

Finally, DeepLense help combat the problem of noisy and low-resolution of real lensing images through various super-resolution techniques. Denoising and upscaling of lensing images can help make their study more accurate.

3.3.1 Single Image Super-Resolution with Diffusion Models

Atal Gupta achieves super-resolution of the real-galaxy lensing dataset, in their GSoC 2024 project, Model 4 using a variety of Diffusion Models (DDPM, SR3, SRDiff, ResShift and CG-DPM).

3.3.2 Super-Resolution for Strong Gravitational Lensing

Pranath Reddy performs a comparative study of the super-resolution of strong lensing images in their GSoC 2023 project, using Residual Models with Content Loss and Conditional Diffusion Models, on the Model 1 dataset.

3.3.3 Physics-Informed Unsupervised Super-Resolution of Strong Lensing Images

Anirudh Shankar explores the unsupervised super-resolution of strong lensing images through a Physics-Informed approach in his GSoC 2024 project, built to handle sparse datasets. They use custom datasets using different lens models and light profiles.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages