Applications are invited for a Master’s Student to work in collaboration with the Institute of Space Sciences and Astronomy on the project STRADA – “Self-supervised Transformers for Radio Astronomy Discovery Algorithms”. The student will join the supervisory team, which will be composed of principal investigator Dr Andrea DeMarco (main supervisor) and Dr Simone Riggi (co-supervisor, INAF). The M.Sc. student will be involved in the testing and evaluation of radio-astronomy specific fine-tuned models for downstream tasks on datasets collated by ISSA and INAF.
In the expanding field of radio astronomy, the exponential growth in data generated by radio telescopes presents both a significant challenge and a remarkable opportunity for astronomical discoveries. Traditional machine learning methods, while effective, often require extensive labelled datasets for training, a resource that is scarce and expensive to produce in the domain of astronomy.
The STRADA project, led by Dr Andrea DeMarco, seeks to address this challenge by leveraging self-supervised learning with transformer models, an approach that promises to unlock the full potential of the vast, yet largely untapped, datasets in radio astronomy. Self-supervised learning, a subset of unsupervised learning techniques, enables models to learn rich representations of data without the need for explicit labels, by predicting parts of the input from other parts. When combined with the powerful transformer architecture, known for its success in understanding complex patterns in data, this approach offers a new pathway for the automated detection and classification of astronomical phenomena. By doing so, STRADA aims to significantly enhance the capability to discover and study transient events, pulsars, and other celestial sources in the radio spectrum, which are key to understanding the universe's most energetic processes. The scientific basis of STRADA lies in its innovative use of self-supervised transformers to interpret the inherently complex and noisy data produced by radio telescopes.
This project will not only develop novel algorithms for data analysis but also contribute to the field of machine learning by adapting and refining transformer models for the unique challenges of radio astronomical data. The anticipated outcome is a set of robust, adaptable tools that can efficiently process large volumes of data, reduce the reliance on labelled datasets, and accelerate the pace of discovery in radio astronomy.
The STRADA project, led by Dr Andrea DeMarco, seeks to address this challenge by leveraging self-supervised learning with transformer models, an approach that promises to unlock the full potential of the vast, yet largely untapped, datasets in radio astronomy. Self-supervised learning, a subset of unsupervised learning techniques, enables models to learn rich representations of data without the need for explicit labels, by predicting parts of the input from other parts. When combined with the powerful transformer architecture, known for its success in understanding complex patterns in data, this approach offers a new pathway for the automated detection and classification of astronomical phenomena. By doing so, STRADA aims to significantly enhance the capability to discover and study transient events, pulsars, and other celestial sources in the radio spectrum, which are key to understanding the universe's most energetic processes. The scientific basis of STRADA lies in its innovative use of self-supervised transformers to interpret the inherently complex and noisy data produced by radio telescopes.
This project will not only develop novel algorithms for data analysis but also contribute to the field of machine learning by adapting and refining transformer models for the unique challenges of radio astronomical data. The anticipated outcome is a set of robust, adaptable tools that can efficiently process large volumes of data, reduce the reliance on labelled datasets, and accelerate the pace of discovery in radio astronomy.
- Applicants must be in possession of an undergraduate degree (Second Class Upper Division or higher) in a STEM subject, particularly in fields such as Artificial Intelligence, Physics, Statistics and Operations Research and Computer Science. Preference will be given to candidates providing evidence of proficiency with the Python programming language, have a good mathematical background, and have a solid understanding of machine learning system design and evaluation. Students who are currently in their final year of the undergraduate programme will also be considered, subject to successful completion of their programme of studies in line with the requirements outlined here.
- The scholarship has a duration of one year, with a further extension of 6 months. The selected candidate will be required to start their studies on a full-time basis in February 2025 and will be receiving a stipend of EUR 500 per month for one year. The tuition fee for the first year of studies for local candidates (up to EUR 400) will be covered from the STRADA project. Further information on applicable fees are available online.
- In case of either extension or suspension of studies, the candidate shall still receive a stipend of €500 per month, provided that i) the maximum allocated funds to the scholarship from the project are not exceeded and ii) the timeframe of the studies is still within the duration of the STRADA project.
- The Scholarship will be subject to the eligibility of the candidates to meet the entry requirements set out by the University of Malta and the conditions for enrolling in the Masters by research programme.
- International applicants are required to present an internationally recognized English Language proficiency Certificate at the required level. Further details are available online. International applicants are also advised to familiarise themselves with information available on the website of the International Office, particularly to the sections concerning visas and e-residence permit conditions and requirements. International candidates will need to provide assurance that they have the funds available to cover their stay in Malta. The stipend awarded for this scholarship is to be considered as a form of allowance and will not cover all the costs attached to studying and living in Malta.
- The degree has an annual enrolment fee of EUR 400 for local/EU/EEA applicants and total tuition fees of EUR 13,400 for non-EU/non-EEA applicants. The University of Malta is an equal opportunity employer and seeks to attract top quality students from countries outside the European Union/EEA. Following acceptance to read for the MSc (on a full-time basis), non-EU/EEA applicants may apply for a tuition fee waiver scholarship. Following scrutiny of the application, deserving students may enjoy a tuition fee waiver of 40% to 100%. The fee waiver will apply for the normal duration of the programme being followed. If extensions of the period of study are required, these may not necessarily be eligible for a fee waiver. Only applicants who have been accepted unconditionally to read for the Master by research on a full-time basis are eligible to apply. Getting a fee waiver is not an automatic right for the selected candidate. Further information on fee waiver scholarships is available online.
- Candidates should submit their letter of application, a copy of their curriculum vitae, copies of their certificates and contact details (including email address) of 2 referees. Applications may be sent by email. Applications should be received by not later than 1 December 2024. International candidates are required to present certified translations of their qualifications into English. Late applications will not be considered.
Note: The scholarship awardee will be required to complete a Master’s by Research application, which is considered separately.
- The selection procedure will involve:
a. scrutiny of qualifications and experience claimed, supported by testimonials and/or certificates;
b. shortlisting; and
c. an interview and/or extended interview.
- Further information may be obtained by contacting Dr Andrea DeMarco by email.
