Abstract Detail



The Potential of Machine Learning for Plant Biology

Soltis, Pamela [1], Nelson, Gil [2].

The Potential of Machine Learning for Plant Biology.

Synopsis Machine learning approaches are highly promising technologies to help address a range of scientific questions in plant science. For example, deep learning technologies have recently achieved impressive performance on a variety of predictive tasks, such as species identification, plant trait recognition, plant species distribution modelling, weed detection, and mercury damage to herbarium specimens. They are also being applied to questions of comparative genomics and gene expression and to conduct high-throughput phenotyping for agricultural and ecological research. Moreover, novel approaches are poised to revolutionize studies of plant phenology and functional traits through application to more than 22 million images of herbarium specimens now available at iDigBio (www.idigbio.org) as well as other digital repositories. As an example, extensive attempts to use deep learning to tackle the difficult taxonomic task of identifying species in large collections of herbarium specimens showed that convolutional neural networks trained on thousands of digitized herbarium sheets are able to learn highly discriminative patterns. These results are very promising for extracting a broad range of accurate annotations in a fully automated way. For example, such approaches could also be used to identify plant phenophase (important for assessing the effects of climate change on plant growth and reproduction and for comparing plant responses with those of pollinators, migratory birds, and other species that rely on plants for food and/or nesting sites) or to extract other evolutionary or ecological traits, such as leaf shape and size, leaf margins, and flower color, to name a few. However, despite the promise of applying deep learning to herbarium specimen images to address a range of questions, this emerging field also raises challenging methodological questions about how to avoid any bias and misleading conclusions when analyzing the produced data. Indeed, as for any statistical learning method, convolutional neural networks are sensitive to bias issues, including the way in which the training datasets are built. Moreover, as good as the prediction might be on average, the quality of the produced annotations can be very heterogeneous from one sample to another, depending on various factors such as the morphology of the species, the storage conditions in which the specimen was preserved, the age of the specimen, etc. Given both the opportunities and challenges, additional research into this topic is needed to enable greater applicability to a broad range of scientific questions. Relevance: What are the potential and limitations of machine learning for plant biology? The field of machine learning is moving rapidly, with the development of alternative approaches that may be best suited to specific questions, data sources, and analytical techniques. In this symposium, we will provide an introduction to the field of machine learning and its potential for plant biology. Speakers will then describe progress, challenges, and solutions for use of machine learning techniques for species identification (from photographs of plants in nature and images of herbarium specimens), analyses of phenology and morphological/functional traits, assessment of damage to herbarium specimens, and stress phenotyping. A special issue of Applications in Plant Sciences will publish papers resulting from this symposium. Partial funding will be provided by iDigBio and the University of Florida Biodiversity Institute. Speakers represent three countries, include two women and four men, range from post-doc to professor, and are all confirmed.


1 - University Of Florida, Florida Museum Of Natural History, Po Box 117800, Gainesville, FL, 32611, United States
2 - IDigBio/Florida State University, 157 Leonards Dr., Thomasville, GA, 31792, United States

Keywords:
machine learning
deep learning
convolutional neural networks
herbarium specimen images
phenology
phenotyping
cyberinfrastructure.

Presentation Type: Symposium Presentation
Number:
Abstract ID:34
Candidate for Awards:None


Copyright © 2000-2019, Botanical Society of America. All rights reserved