{
  "annotation_id": "d55a9ca3-f166-4ff8-a3eb-287f15bd48a4",
  "date_created": "2026-03-02T18:01:35.307000Z",
  "date_modified": "2026-03-02T18:01:35.307000Z",
  "file_hash": "37f4e0fedf5df33d51750bfb8145aac365cde0da2fc3bfea033e9f57c1daf474",
  "private": false,
  "record": {
    "abstract": "Tree shape statistics quantify some aspect of the shape of a phylogenetic\ntree. They are commonly used to compare reconstructed trees to evolutionary\nmodels and to find evidence of tree reconstruction bias. Historically, to find\na useful tree shape statistic, formulas have been invented by hand and then\nevaluated for utility. This article presents the first method which is capable\nof optimizing over a class of tree shape statistics, called Binary Recursive\nTree Shape Statistics (BRTSS). After defining the BRTSS class, a set of\nalgebraic expressions is defined which can be used in the recursions. The tree\nshape statistics definable using these expressions in the BRTSS is very\ngeneral, and includes many of the statistics with which phylogenetic\nresearchers are already familiar. We then present a practical genetic algorithm\nwhich is capable of performing optimization over BRTSS given any objective\nfunction. The chapter concludes with a successful application of the methods to\nfind a new statistic which indicates a significant difference between two\ndistributions on trees which were previously postulated to have similar\nproperties.",
    "arxiv_id": "q-bio/0605034",
    "authors": [
      "Frederick A. Matsen"
    ],
    "categories": [
      "q-bio.PE"
    ],
    "title": "Optimization over a class of tree shape statistics",
    "url": "https://arxiv.org/abs/q-bio/0605034"
  },
  "schema_id": "dorsal/arxiv",
  "source": {
    "execution_id": "30e8f1b2-e931-470f-8bcf-c5c13b7f672c",
    "id": "arXiv Dataset IDs",
    "type": "Model",
    "variant": "snapshot-2026-03-01",
    "version": "0.1.0"
  },
  "user_id": 1000002
}