EPSRC Centre for Doctoral Training Cloud Computing for Big Data


Ben Jaeger
profile photo

I graduated from Aberystwyth University with a Data Science BSc in 2020 before joining the CDT. 

I did my undergraduate final project in text mining, attempting to make an AI able to give human-usable and game-legal word association hints for the boardgame Codenames. 

I was attracted to the CDT because of the collaborative element, the opportunity to work and learn with my peers appeals to me greatly. I also look forward to pushing myself with the engaging and challenging content of the course. 

Outside of academia I enjoy playing boardgames, archery and swimming. 

PhD Title

Applying Neural Architecture Search Solutions to Multitasking Problems in Deep Learning

There have been many innovations in finding efficient and effective neural network architectures using automated systems over the years but these techniques have so far largely only been applied in cases where the model is optimising for one task. In this PhD, I hope to try expanding these techniques further to optimise for multiple tasks.

My first project, titled MLFnet, aims to build a network layer by layer (a technique from NAS that goes by a few names, one being Learning Forward) while examining task affinity as the model grows, separating tasks out when they become incompatible, thus adaptively sharing information when it is beneficial but allowing tasks to branch out and continue learning when this is not the case.


Amir Atapour-Abarghouei, Stephen McGough, Matt Forshaw