Abstract Detail

Crops and Wild Relatives

Wizenberg, Sydney Barrett [1], Suhanic, West [1], Campbell, Lesley Geills [2].

How can we breed better weed? Implementing artificial neural networks for predicting yield in Cannabis sativa L.

The ability to model and predict the yield of valuable crops like Cannabis sativa could improve breeding strategies, but the industry is currently limited by the use of rudimentary statistical methods. Machine learning may provide a complex multifaceted modelling environment, but its rarely used in the context of crop breeding, wherein artificial selection often relies on a yield based trait, or phenotypic measurements of a correlated trait. Previous attempts to identify early indicators of yield in C. sativa have had limited success, likely as a result of reliance on elementary statistical methods, such as structural equation modelling. To address this, we developed a framework for implementing artificial neural networks to predict yield (floral biomass, cannabinoid abundance) in C. sativa, by amalgamating machine learning methods with three types of preliminary linear analyses. Through development of this framework we demonstrated that artificial neural networks can successfully predict yield using morphological data, and can identify correlations between agricultural traits. Our work focused on addressing limitations of artificial neural networks, such as model overfitting, and an inability to differentiate between relevant and non-relevant input, by comparing the relative utility of three means of preliminary analyses. Our work has made substantial progress towards understanding how to breed better Cannabis and has shown that plant architecture can predict yield in multiple genotypes and environments.

1 - Ryerson University, Chemistry and Biology , 350 Victoria Street, Toronto, Ontario, M5B 2K3, Canada
2 - Ryerson University, 350 Victoria Street, Toronto, Ontario, M5B 2K3, Canada

Keywords:
Machine learning
Artifical neural networks
Cannabis sativa
Crop science
Crop breeding.

Presentation Type: Oral Paper
Session: CW2, Crops and Wild Relatives II
Location: /
Date: Wednesday, July 21st, 2021
Time: 3:15 PM(EDT)
Number: CW2002
Abstract ID:661
Candidate for Awards:None