Home Framework Framework for improving artificial intelligence in the years to come

Framework for improving artificial intelligence in the years to come

0

Experts from the University of Essex have discovered a radical new framework for improving artificial intelligence (AI) in the coming years.

Image Credit: Shutterstock.com/cono0430

The Essex team hope their research will serve as the foundation for the next generation of advances in AI and machine learning. The study was published in the top machine learning journal, Machine Learning Research Journal.

This could lead to advances in everything from driverless cars and smartphones that understand voice commands to stronger automatic medical diagnostics and drug development.

Artificial intelligence research ultimately aims to produce fully autonomous, intelligent machines that we can converse with and perform tasks for us, and this newly published work accelerates our progress towards that..

Dr Michael Fairbank, Study Co-Lead Author, School of Computing and Electronic Engineering, University of Essex

“Deep learning” – which involves training multi-layered artificial neural networks to solve a task – has been used in the latest remarkable advances in AI around vision tasks, speech recognition and translation software . However, training these deep neural networks is a computationally expensive challenge that requires a large number of training examples as well as computational time.

The Essex team, which includes Professor Luca Citi and Dr Spyros Samothrakis, have developed a completely different approach to coaching deep learning neural networks.

Our new method, which we call Target Space, offers researchers a step change in how they can improve and build their AI creations. Target Space is a paradigm-shifting vision that disrupts the process of forming these deep neural networks, ultimately accelerating current advancements in AI developments..

Dr Michael Fairbank, Study Co-Lead Author, School of Computing and Electronic Engineering, University of Essex

The standard method of training neural networks to increase efficiency is to regularly make small adjustments to the connection strengths between neurons in the network. The Essex team adopted a new strategy. Instead of adjusting the strength of the connections between neurons, the new “target-space” technique proposes adjusting the firing strengths of the neurons themselves.

This new method dramatically stabilizes the learning process, through a process we call cascading disentanglement. This allows the trained neural networks to be deeper, and therefore more capable, and at the same time potentially requiring fewer training examples and fewer computational resources. We hope this work will provide a foundation for the next generation of breakthroughs in artificial intelligence and machine learning..

Luca Citi, Professor, University of Essex

The method will be applied to a variety of new academic and industrial applications in the coming months.

Journal reference:

Fairbank, M. et al. (2022) Deep learning in target space. Machine Learning Research Journal. Available at: https://jmlr.org/papers/v23/20-040.html.

Source: https://www.essex.ac.uk/