Imma Perfetto
Cosmos science journalist
In pursuit of creating artificial intelligence that can “think” more like a person, researchers have developed a new machine learning algorithm that uncovers patterns in data without human guidance.
The algorithm, called Torque Clustering, can efficiently and autonomously analyse vast amounts of data.
It could be used to detect disease patterns, uncover fraud, or understand behaviour if used in fields such as medicine, finance, and psychology. The open-source code has been made available to researchers.
According to Chin-Teng Lin – a distinguished professor at the University of Technology Sydney (UTS) in Australia and co-author of a paper detailing the method – nearly all current artificial intelligence technologies rely on ‘supervised learning’.
“This is an AI training method that requires large amounts of data to be labelled by a human using predefined categories or values, so that the AI can make predictions and see relationships,” he explains.
But supervised learning has a number of limitations.
“Labelling data is costly, time-consuming and often impractical for complex or large-scale tasks,” Lin explains.
Unsupervised learning works with unlabelled data.
“In nature, animals learn by observing, exploring, and interacting with their environment, without explicit instructions,” says Lin.
“The next wave of AI, ‘unsupervised learning’, aims to mimic this approach.”
Clustering is a common technique used in many fields of science, which involves grouping a set of objects together. Objects within a group, or cluster, are more similar to each other than to objects in another cluster.
Lin and lead author of the paper, Dr Jie Yang, also from UTS, found that their new Torque Clustering method outperforms all other state-of-the-art clustering algorithms.
“What sets Torque Clustering apart is its foundation in the physical concept of torque, enabling it to identify clusters autonomously and adapt seamlessly to diverse data types, with varying shapes, densities, and noise degrees,” says Yang.
“It was inspired by the torque balance in gravitational interactions when galaxies merge. It is based on 2 natural properties of the universe: mass and distance.”
The researchers explain in the paper that “…Torque Clustering simulates the process of galaxy minor mergers, so that clusters with larger masses continuously merge adjacent clusters with smaller masses.”
The researchers suggest Torque Clustering could also support the development of general artificial intelligence, particularly in robotics and autonomous systems, by helping to optimise movement, control and decision-making.
The research appears in the journal EEE Transactions on Pattern Analysis and Machine Intelligence.
