Picture of Alberto Gil Ramos

Alberto Gil Ramos

Cambridge, UK
Research Scientist


PhD in Applied Mathematics, University of Cambridge, UK, 2016
Master in Statistics and Probability, University of Porto, Portugal, 2010
First Degree in Mathematics, University of Porto, Portugal, 2007


I have a PhD from the University of Cambridge and over two years industry experience working with image, health and audio data, where I leverage my expertise in machine learning and applied mathematics to understand, model and explain data for informed decision making.

Research Interests

  • Artificial Intelligence
  • Computational & Algorithmic Sciences
  • Deep Learning
  • Machine Learning
  • Statistical & Data Sciences

Honors and Awards

2015 - 2016
London Mathematical Society 150th anniversary grant;
Funded by London Mathematical Society

2011 - 2015
Doctoral fellowship;
Funded by Portuguese Foundation for Science and Technology

Design of next generation very high data rate wireless communication systems;
Funded by Portuguese Foundation for Science and Technology

2009 - 2010
Analysis, design and optimisation of future communication systems;
Funded by Institute of Telecommunications and Portuguese Foundation for Science and Technology

2008 - 2009
Design and optimisation of wavelength-division multiplexing millimetre-wave fibre-radio systems;
Funded by Portuguese Foundation for Science and Technology

Selected Articles and Publications

Submitted papers:

Gordon, Gataric, R., Mouthaan, Williams, Yoon, Wilkinson and Bohndiek
arXiv:1904.02644 "Characterising optical fibre transmission matrices using metasurface reflector stacks for lensless imaging without distal access"

Published journal papers:

Gataric, Gordon, Renna, R., Alcolea and Bohndiek
IEEE Transactions on Medical Imaging, DOI:10.1109/TMI.2018.2875875 "Reconstruction of optical vector-fields with applications in endoscopic imaging"

Springer Advances in Computational Mathematics, DOI:10.1007/s10444-017-9547-7 "Uniform and high-order discretisation schemes for Sturm--Liouville problems via Fer streamers"

R. and Iserles
Springer Numerische Mathematik, DOI:10.1007/s00211-014-0695-0 "Numerical solution of Sturm--Liouville problems via Fer streamers"

R. and Rodrigues 
IEEE Transactions on Information Theory, DOI:10.1109/TIT.2014.2333746 "Fading channels with arbitrary inputs: asymptotics of the constrained capacity and information and estimation measures"

Published conference papers:

Gong, R., Mathur, Bhattacharya and Kawsar
IEEE International Conference on Machine Learning and Applications, to appear; "AudiDoS: Real-Time Denial-of-Service Adversarial Attacks on Deep Audio Models"

R. and Rodrigues
IEEE International Symposium on Information Theory, DOI:10.1109/ISIT.2013.6620592 "Characterisation and optimisation of the constrained capacity of coherent fading channels driven by arbitrary inputs: a Mellin transform based asymptotic approach"

Rodrigues and R.
IEEE International Symposium on Information Theory, DOI:10.1109/ISIT.2009.5206051 "On multiple-input multiple-output Gaussian channels with arbitrary inputs subject to jamming"

Published workshop papers:

Bhattacharya, R., Kawsar, Lane, Gionta, Manidis, Silvesti and Vegreville
Ubiactivity Workshop, ACM International Joint Conference on Pervasive and Ubiquitous Computing, DOI:10.1145/3267305.3267682  "Monitoring daily activities of multiple sclerosis patients with connected health devices"


Filed patents:

R. and Bhattacharya
Nokia Technologies Oy; NC105910, EP; 18188739.9; "Practical defences for audio adversarial attacks via audio pre-processing for systems with two or more microphones within the same or across different physical devices through subsampling and fusion of misaligned audio time series"

Bhattacharya and R.
Nokia Technologies Oy; NC105898, EP; 18183849.1; "Making deep learning model compression robust to adversarial audio perturbations"

R., Bhattacharya, Kawsar and Lane
Nokia Technologies Oy; NC103649, EP; 17208716.5; "Decentralized household training of family and individual health/management predictive learning models across heterogeneous constrained devices with built-in user anonymity and accumulative local training batches"

Baykaner and R.
Nokia Technologies Oy; PCT/EP2017/053682; WO2018149504A1; "Changing smart contracts recorded in block chains"