Machine Learning Week Europe is live in Berlin! But what even is it? Machine learning is a form of artificial intelligence, where software applications and computers learn how to improve and redesign their ability to do tasks – completely independently from their human creators. When commercially deployed, it’s called predictive analytics. Around the world, these tools are being exponentially rolled-out for a multitude of applications, such as big data management, speech recognition and analytical model building. It’s seen as the future of computing and will be at the forefront of future discoveries. Delve into our roundup of the latest research below and familiarise yourself with humanity’s new best friend.
Genome sequencing platforms are transforming the field of genetic disease research as they offer a closer look at human genes and DNA for clinical diagnostics. Based on machine learning methods, Deep Simulator provides simulated datasets to train and test sequencing analytical tools while WaveNano has innovated the process of translating a raw signal sequence into a DNA read. Dr Xin Gao is a professor of computer science in CEMSE Division at King Abdullah University of Science and Technology (KAUST), Associate Director of Computational Bioscience Research Center, and Deputy Director of Smart Health Initiative at KAUST. With his team he studies the intersections between computer science and biology/biomedicine to develop novel genome sequencing methods.
Machine learning has become a staple of research into many of today’s most cutting-edge technologies. Until now, however, it has not been widely considered as a useful tool for online optimisation of the performance of particle accelerators. Through his research, Dr Alexander Scheinker at the Los Alamos National Laboratory in New Mexico, USA, introduces new techniques based on machine learning, which can adaptively tune the bewilderingly complex controls of particle accelerators in real time without human input. His team’s work could ultimately make it far easier for researchers across many different academic fields to make ground-breaking new discoveries.
Shuichiro Makigaki and DrTakashi Ishida, from the Department of Computer Science at Tokyo Institute of Technology, are developing a new sequence alignment generation model that employs machine learning and dynamic programming to predict protein structures. This novel methodology can also be applied to homology detection which is fundamental to bioinformatics.
Dr Daisuke Kihara’s team at Purdue University have created novel computational approaches for predicting protein functions. Instead of following a one-protein-one-function approach, their algorithms can predict the functional relationships of entire groups of proteins related to a specific biological process. The team has also expanded into mining oversighted or previously unknown proteins that have multiple, independent functions. The team’s methods challenge the logic behind conventional protein function studies and propose tools that may better capture the complicated nature of protein interactions in biological processes.
Much of what we know about how our universe works has been learnt by analysing the astronomical signals captured from the sky. However, these signals will inevitably have some noise associated with them – so how can astronomers be sure that their observations of strange, unexpected signals reflect reality? Edward Higson at the University of Cambridge and his colleagues have made progress in dealing with this issue, by constructing machine learning algorithms which can process noisy astronomical images to reconstruct clear ones. Using
Bayesian statistics, their software provides astronomers with useful tools for processing their observations.