Adding the KBest Algorithm

ANN-Benchmarks contains many algorithms. You can also use Huawei-developed algorithm KBest to search for datasets. To add the KBest algorithm, perform the following steps:

1. Add the implementation of the KBest algorithm.
   1. Open the module.py file.

      1	vim /data/ann-benchmarks-main/ann_benchmarks/algorithms/milvus/module.py

   2. Add the following content at the end of the file:

      1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

      class MilvusKBEST(Milvus): def __init__(self, metric, dim, index_param): super().__init__(metric, dim, index_param) self._index_R = index_param.get("R", None) self._index_L = index_param.get("L", None) self._index_A = index_param.get("A", None) self._index_btype = index_param.get("init_builder_type", None) self._index_consecutive = index_param.get("consecutive", None) self._index_level = index_param.get("level", None) self._build_index_type = index_param.get("build_index_type", None) self._graph_opt_iter = index_param.get("graph_opt_iter", None) self._reorder = index_param.get("reorder", None) def get_index_param(self): return { "index_type": "KBEST", "params": { "R": self._index_R, "L": self._index_L, "A": self._index_A, "init_builder_type": self._index_btype, "consecutive": self._index_consecutive, "level": self._index_level, "build_index_type": self._build_index_type, "graph_opt_iter": self._graph_opt_iter, "reorder": self._reorder }, "metric_type": self._metric_type } def set_query_arguments(self, ef, num_threads, adding_pref, patience): self.search_params = { "metric_type": self._metric_type, "params": { "efs": ef, "num_search_thread": num_threads, "adding_pref": adding_pref, "patience": patience } } self.name = f"MilvusKBEST metric:{self._metric}, index_R:{self._index_R}, index_L:{self._index_L}, index_A:{self._index_A}, index_btype:{self._index_btype}, index_consecutive:{self._index_consecutive}, index_level:{self._index_level}, build_type={self._build_index_type}, graph_iter={self._graph_opt_iter}, reorder={self._reorder}, search_ef:{ef}, search_thread={num_threads}, adding_pref={adding_pref}, patience={patience}"

2. Add the KBest algorithm configuration.
   1. Open the config.yaml file.

      1	vim ann_benchmarks/algorithms/milvus/config.yaml

   2. Add the following content at the end of the file:

      1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

      - base_args: ["@metric", "@dimension"] constructor: MilvusKBEST disabled: false docker_tag: ann-benchmarks-milvus module: ann_benchmarks.algorithms.milvus name: milvus-kbest run_groups: KBest: args: R: [50] L: [100] A: [60] init_builder_type: ["RNNDescent"] consecutive: [20] level: [2] build_index_type: ["SSG"] graph_opt_iter: [6] reorder: [true] query_args: [[400], [1], [52], [80]]

Table 1 KBest parameters describes the KBest parameters. The reference values are determined based the query result precision, memory consumption, and time consumption. You can set the parameters as required.