Machine learning opens new doors in archeology

Human figure rock art created over thousands of years in Arnhem Land, Australia has been subjected to a transformative machine learning study to analyze changes in style over the years.

The study tested different styles called “Nordic figures”, “dynamic figures”, “post-dynamic figures” and “simple figures with boomerangs” to understand how these styles relate to each other.

Working with traditional owners Mimal and Marrku of the Wilton River region in Australia’s high end, Southern Australian researchers led by Flinders University archaeologist Dr Daryl Wesley further examined near the art of this region.

Jarrad Kowlessar, researcher at Flinders, and the team used machine learning to analyze rock art images collected during surveys conducted in the land of Marrku in 2018 and 2019.

Co-authors include Dudley Lawrence and Abraham Weson and others from the Mimal Land Management Aboriginal Corporation, Alfred Nayinggul from the Njanjma Aboriginal Corporation, Dr Ian Moffat from Flinders and University of Adelaide researcher James Keal.

The reconstructed rock art timeline, which was just published in Australian Archeology, uses existing datasets of over 14 million different photos of a wide range of things, from animals such as dogs, cats , lizards and insects to objects such as chairs, tables and cups.

“In total, the computer has seen over 1000 different types of objects and learned to tell the difference between them just by looking at pictures of them,” says Dr Wesley.

“The important skill that this computer developed was a mathematical model that has the ability to tell how similar two different images are to each other.

Next, mathematical modeling was applied to the images collected in northern Australia.

“This approach allows us to use the computer program to show how unique rock art is in the Wilton River and how it relates to rock art in other parts of Arnhem Land,” says Dr. Wesley.

“We can use this to help show how rock art styles are shared by the traditional owners of Arnhem Land and which are unique to each group throughout the past.”

Machine learning allows a computer to “ learn ” different things about information that can take a human many years to navigate and learn, says a PhD from Flinders University. archaeological candidate Jarrad Kowlessar, pioneer of the machine learning approach to rock art analysis.

“One amazing result is that the machine learning approach ordered the styles in the same timeline that archaeologists ordered them by inspecting which ones appear above which. This shows that similarity and time are closely related in Arnhem Land rock art and human figures close in time were more similar to each other than those drawn for a long time, ”he says.

“For example, the machine learning algorithm plotted Northern Running digits and dynamic digits very close to each other on the graph it produces. This shows that these styles that we know are closer to each other by ages are also closer to each other in appearance, which could be a very difficult thing to notice without an approach like this ”.

The article points out that the new methodology eliminated a large degree of individual human interpretation and possible bias by using a machine learning approach called “transfer learning”.

This allowed the computer to understand how each style relates directly to each other, regardless of the researchers involved.

Researchers are excited about this methodology which opens up new perspectives for a large amount of archaeological research to understand all kinds of different human material cultures in a different way.