While investigating the mysterious teeth, machine learning models were used to identify the oldest therizinosaur fossils ever discovered in Britain. Therizinosaurs were first identified in fossils discovered in Great Britain thanks to toothy remains found in Oxfordshire, Gloucestershire and Dorset.
Researchers from the Natural History Museum and Birkbeck College trained computer models to recognize mysterious teeth using advanced machine learning techniques that trace the origins of some members of the group back nearly 30 million years, NHM reported.
“Previous studies have suggested that the manipulators were in the middle Jurassic period, but the actual fossil record was unclear and contradictory. Along with fossils found elsewhere, this study shows that the group had reached a global distribution by that time,” said Simon Wills, a doctoral student at the Museum of Natural History, who led the research. ” said.
“The teeth we analyzed include the only troodontid and therizinosaur fossils ever recorded in the UK and are the oldest evidence of these dinosaurs anywhere in the world.”
Machine learning models can recognize isolated teeth
A large herbivorous dinosaur of the Late Cretaceous, Therizinosaurus was distinguished by its long claws resembling scissor blades. These prehistoric species were featured in the latest Jurassic World movie due to their distinctive appearance.
Although previous studies have tried to classify isolated teeth using several statistical methods, this has not always been successful. The authors of the current study sought to reinforce this after showing that machine learning models could recognize individual teeth of known species with high accuracy.
“The use of machine learning in vertebrate paleontology is still in its infancy, although its use is increasing,” says Simon.
“The main disadvantage is the need for a comprehensive training dataset to train the models. We were fortunate in our research to have a relatively large dataset of measurements of dinosaur teeth that we can use to train the models,” adds Simon.