Novel applications of machine learning in cosmology and beyond
Lead Research Organisation:
University College London
Department Name: Physics and Astronomy
Abstract
With current surveys such as the Dark Energy Survey (DES), and in anticipation of even bigger projects like LSST and SKQ, the field of Astronomy is currently transitioning into an era dominated by enormous datasets. Indeed there will soon be more data than it would ever be possible to analyse, and although this influx of data could allow for a golden age of discovery, there is also the risk that the most interesting discoveries could be lost, drowned in the amount of data, never to be found. Machine learning is probably the best way we can address this problem, and in this project we aim to apply various machine learning algorithms to DES data to help analyse the data more efficiently and fuel discoveries
Studentship Projects
| Project Reference | Relationship | Related To | Start | End | Student Name |
|---|---|---|---|---|---|
| ST/P006736/1 | 01/10/2017 | 30/09/2024 | |||
| 1966844 | Studentship | ST/P006736/1 | 01/10/2017 | 30/09/2021 | Ben Henghes |
| Description | ASOS |
| Organisation | ASOS |
| Country | United Kingdom |
| Sector | Private |
| PI Contribution | Was a member of a group project to create a recommendation system which could predict what sizes a customer should purchase. Also about to start a 6-month placement working on forecasting sales for Black Friday to set the prices of every item on the website. |
| Collaborator Contribution | Made the project and provided the data as well as some training to be able to quickly understand the dataframes. |
| Impact | Finished the group project with a good result of being able to improve the current method used to provide recommendations for shoe sizes. |
| Start Year | 2018 |
| Description | STFC |
| Organisation | Science and Technologies Facilities Council (STFC) |
| Country | United Kingdom |
| Sector | Public |
| PI Contribution | Working on implementing machine learning for photometric redshift estimation and performing benchmarks to compare the efficiencies of different methods. We provided more of the astronomy background and writing up the work into a journal article. |
| Collaborator Contribution | Came up with the idea to use benchmarking as a way of having a new element to this project and ran through the machine learning algorithms on SDSS data to get redshift estimations. |
| Impact | We are still in the process of writing up the results into a paper. |
| Start Year | 2019 |
| Description | University of Michigan |
| Organisation | University of Michigan |
| Country | United States |
| Sector | Academic/University |
| PI Contribution | Working to use machine learning to improve the TNO detection pipeline which was being run by the group at Michigan to search DES data. |
| Collaborator Contribution | They provided the code which was being used and that we were trying to improve, as well as the training data. |
| Impact | We are currently in the process of publishing a paper (still in internal review) |
| Start Year | 2017 |