US Researchers* published in the Proceedings of the National Academy of Sciences have applied machine learning to the identification of fossil pollen grains. The team developed and trained three machine-learning models to differentiate between several existing Amherstieae legume genera and tested them against fossil specimens from western Africa and northern South America dating back to the Paleocene, Eocene and Miocene.
Taxonomic resolution is said to be a major challenge in palynology that limits the ecological and evolutionary interpretations possible with deep-time fossil pollen data. The NIST-backed team used optical ‘super resolution’ microscopy and machine learning to create a quantitative workflow for producing palynological identifications and hypotheses of biological affinity. Three convolutional neural network classification models were developed: maximum projection, multi-slice, and fused. After training on modern genera, the models were run on the fossils. All models achieved average accuracies of 83 to 90% in the classification of the extant genera. The majority (86%) of fossil identifications showed consensus among at least two of the three models. The study supported the hypothesis that Amherstieae originated in Paleocene Africa and dispersed to South America during the Paleocene-Eocene thermal maximum (56 Ma).
While one might debate the exact meaning of machine learning and AI, using the computer to classify fossils has a long history. Earlier work, termed ‘numerical taxonomy’ dates back to the 1960s. See for instance, John Schrock’s intriguing book review on Amazon.
* A US National Science Foundation-funded team at the Smithsonian Tropical Research Institute, the University of Illinois at Urbana-Champaign, the University of California, Irvine and collaborating institutions.
© Oil IT Journal - all rights reserved.