Species determination using AI machine-learning algorithms: Hebeloma as a case study

IMA Fungus

Table 3 Species identification success metrics for various choices of characters used to calibrate the classifier

In these results, neither the continent filter nor second pass is applied, and collections without recorded data for at most one character are permitted. The optimization method is AdamW and the network shape is one hidden layer with a ReLU activation function. In this and subsequent results tables, the final five columns refer to the Top 1, Top 3, Top 5, Mean Reciprocal Rank and macro F1 metrics respectively, all scaled to give a score out of 100, which would represent perfect prediction

ISSN: 2210-6359