Alasdair Paren

Machine Learning Reseacher

DPhil Candidate - AIMS Oxford

Position

Research Interests

My research focuses on efficient optimisation of neural networks. Specifically the efficient optimisation algorithms that produce high accuracy models with little hyperparameter cross validation, and the training of efficient quantised neural networks. While much of modern machine learning research leverages larger and larger models, it is important to remember the real world energy costs associated with training and deploying such architectures. I believe it is important to focus on reducing the cost of machine learning to help maximise its use cases and reduce its impact on the planet.

Focus

Optimisation

Quantized Neural Networks

Computer Vision

Publications

Papers

A Stochastic Bundle Method for Interpolating Networks

JMLR 2021 - available here

Paren, A., Berrada, L., Poudel, R., Kumar, P.

Faking Interpolation Until You Make It

TMLR 2022 - available here

Paren, A., Poudel, R., Kumar, P.

Training Fully Binary Neural Networks the Easy Way

BMVC 2022

Paren, A., R., Poudel

Experience

Profile

Computer Vision Group Reseach Intern

Toshiba

2021-2022

Graduate Engineer

WSP

2014 - 2016

Education

Academics

DPhil - Optimiation for Machine Learning

University of Oxford

2017 - 2022

MSc - Computational Statistics and Machine Learning

University College London

2016 - 2017

MEng - Mechanical Engineering

Imperial College London

2010 - 2014

Tech I'm familiar with

Langages

Python

Pytorch

Git

Latex

Matlab

SolidWorks

Machine Tools

Keen to learn more

Get in touch

Contact

I'm best contacted via email at alasdair.paren@gmail.com