Clustering.yaml

id: clustering_example
description: Clustering benchmark on Gagolewski's, true number of clusters plus minus 2.
version: 1.2
benchmarker: "Izaskun Mallona, Daniel Incicau"
storage: https://play.min.io
benchmark_yaml_spec: 0.04
storage_api: S3
storage_bucket_name: clustering_example
software_backend: conda
software_environments:
  clustbench:
    description: "clustbench on py3.12.6"
    conda: envs/clustbench.yml
    envmodule: clustbench
    apptainer: envs/clustbench.sif
  sklearn:
    description: "Daniel's on py3.12.6"
    conda: envs/sklearn.yml
    apptainer: envs/sklearn.sif
    envmodule: clustbench # not true, but
  R:
    description: "Daniel's R with readr, dplyr, mclust, caret"
    conda: envs/r.yml
    apptainer: envs/r.sif
    envmodule: fcps # not true, but
  rmarkdown:
    description: "R with some plotting dependencies"
    conda: envs/rmarkdown.yml
    apptainer: envs/r.sif # not true, but
    envmodule: fcps # not true, but
  fcps:
    description: "CRAN's FCPS"
    conda: envs/fcps.yml
    apptainer: envs/fcps.sif
    envmodule: fcps
metric_collectors:
  - id: plotting
    name: "Single-backend metric collector."
    software_environment: "rmarkdown"
    repository:
      url: https://github.com/imallona/clustering_report
      commit: 1d6bdf5
    inputs:
      - metrics.scores
    outputs:
      - id: plotting.html
        path: "{input}/{name}/plotting_report.html"

stages:
  ## clustbench data ##########################################################

  - id: data
    modules:
      - id: clustbench
        name: "clustbench datasets, from https://www.sciencedirect.com/science/article/pii/S0020025521010082#t0005 Table1"
        software_environment: "clustbench"
        repository:
          url: https://github.com/imallona/clustbench_data
          commit: 366c5a2
        parameters: # comments depict the possible cardinalities and the number of curated labelsets
          - values: ["--dataset_generator", "fcps", "--dataset_name", "atom"] # 2 1
          # - values: ["--dataset_generator", "fcps", "--dataset_name", "chainlink"] #  2 1
          # - values: ["--dataset_generator", "fcps", "--dataset_name", "engytime"] # 2 2
          # - values: ["--dataset_generator", "fcps", "--dataset_name", "hepta"] #  7 1
          # - values: ["--dataset_generator", "fcps", "--dataset_name", "lsun"] # 3 1
          # - values: ["--dataset_generator", "fcps", "--dataset_name", "target"] # 2, 6  2
          # - values: ["--dataset_generator", "fcps", "--dataset_name", "tetra"] #  4 1
          # - values: ["--dataset_generator", "fcps", "--dataset_name", "twodiamonds"] #  2 1
          # - values: ["--dataset_generator", "fcps", "--dataset_name", "wingnut"] #  2 1
          # - values: ["--dataset_generator", "graves", "--dataset_name", "dense"] #  2 1
          # - values: ["--dataset_generator", "graves", "--dataset_name", "fuzzyx"] # 2, 4, 5 6
          # - values: ["--dataset_generator", "graves", "--dataset_name", "line"] # 2 1
          # - values: ["--dataset_generator", "graves", "--dataset_name", "parabolic"] #  2, 4  2
          # - values: ["--dataset_generator", "graves", "--dataset_name", "ring"] # 2 1
          # - values: ["--dataset_generator", "graves", "--dataset_name", "ring_noisy"] # 2 1
          # - values: ["--dataset_generator", "graves", "--dataset_name", "ring_outliers"] #  2, 5  2
          # - values: ["--dataset_generator", "graves", "--dataset_name", "zigzag"] # 3, 5  2
          # - values: ["--dataset_generator", "graves", "--dataset_name", "zigzag_noisy"] # 3, 5  2
          # - values: ["--dataset_generator", "graves", "--dataset_name", "zigzag_outliers"] #  3, 5  2
          # - values: ["--dataset_generator", "other", "--dataset_name", "chameleon_t4_8k"] # 6 1
          # - values: ["--dataset_generator", "other", "--dataset_name", "chameleon_t5_8k"] # 6 1
          # - values: ["--dataset_generator", "other", "--dataset_name", "hdbscan"] # 6 1
          # - values: ["--dataset_generator", "other", "--dataset_name", "iris"] #  3 1
          # - values: ["--dataset_generator", "other", "--dataset_name", "iris5"] # 3 1
          # - values: ["--dataset_generator", "other", "--dataset_name", "square"] #  2 1
          # - values: ["--dataset_generator", "sipu", "--dataset_name", "aggregation"] #  7 1
          # - values: ["--dataset_generator", "sipu", "--dataset_name", "compound"] # 4, 5, 6 5
          # - values: ["--dataset_generator", "sipu", "--dataset_name", "flame"] #  2 2
          # - values: ["--dataset_generator", "sipu", "--dataset_name", "jain"] # 2 1
          # - values: ["--dataset_generator", "sipu", "--dataset_name", "pathbased"] #  3, 4  2
          # - values: ["--dataset_generator", "sipu", "--dataset_name", "r15"] #  8, 9, 15  3
          # - values: ["--dataset_generator", "sipu", "--dataset_name", "spiral"] # 3 1
          # - values: ["--dataset_generator", "sipu", "--dataset_name", "unbalance"] #  8 1
          # - values: ["--dataset_generator", "uci", "--dataset_name", "ecoli"] # 8 1
          # - values: ["--dataset_generator", "uci", "--dataset_name", "ionosphere"] #  2 1
          # - values: ["--dataset_generator", "uci", "--dataset_name", "sonar"] # 2 1
          # - values: ["--dataset_generator", "uci", "--dataset_name", "statlog"] # 7 1
          # - values: ["--dataset_generator", "uci", "--dataset_name", "wdbc"] #  2 1
          # - values: ["--dataset_generator", "uci", "--dataset_name", "wine"] #  3 1
          # - values: ["--dataset_generator", "uci", "--dataset_name", "yeast"] # 10  1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "circles"] # 4 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "cross"] # 4 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "graph"] # 10  1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "isolation"] # 3 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "labirynth"] # 6 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "mk1"] # 3 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "mk2"] # 2 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "mk3"] # 3 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "mk4"] # 3 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "olympic"] # 5 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "smile"] # 4, 6  2
          # - values: ["--dataset_generator", "wut", "--dataset_name", "stripes"] # 2 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "trajectories"] #  4 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "trapped_lovers"] #  3 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "twosplashes"] # 2 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "windows"] # 5 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "x1"] #  3 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "x2"] #  3 1
          - values: ["--dataset_generator", "wut", "--dataset_name", "x3"] #  4 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "z1"] #  3 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "z2"] #  5 1
          # - values: ["--dataset_generator", "wut", "--dataset_name", "z3"] #  4 1
    outputs:
      - id: data.matrix
        path: "{input}/{stage}/{module}/{params}/{dataset}.data.gz"
      - id: data.true_labels
        path: "{input}/{stage}/{module}/{params}/{dataset}.labels0.gz"

  ## clustbench methods (fastcluster) ###################################################################
  
  - id: clustering
    modules:
      - id: fastcluster
        name: "fastcluster algorithm"
        software_environment: "clustbench"
        repository:
          url: https://github.com/imallona/clustbench_fastcluster
          # url: /home/imallona/src/clustbench_fastcluster/
          commit: "45e43d3"
        parameters:
          - values: ["--linkage", "complete"]
          - values: ["--linkage", "ward"]
          # - values: ["--linkage", "average"]
          # - values: ["--linkage", "weighted"]
          # - values: ["--linkage", "median"]
          # - values: ["--linkage", "centroid"]
      - id: sklearn
        name: "sklearn"
        software_environment: "clustbench"
        repository:
          url: https://github.com/imallona/clustbench_sklearn
          #url: /home/imallona/src/clustbench_sklearn
          commit: 5877378
        parameters:
          - values: ["--method", "birch"]
          - values: ["--method", "kmeans"]
          # - values: ["--method", "spectral"] ## too slow
          # - values: ["--method", "gm"]
      - id: agglomerative
        name: "agglomerative"
        software_environment: "clustbench"
        repository:
          url: https://github.com/imallona/clustbench_agglomerative
          commit: 5454368
        parameters:
          # - values: ["--linkage", "average"]
          - values: ["--linkage", "complete"]
          - values: ["--linkage", "ward"]
      - id: genieclust
        name: "genieclust"
        software_environment: "clustbench"
        repository:
          url: https://github.com/imallona/clustbench_genieclust
          commit: 6090043
        parameters:
          - values: ["--method", "genie", "--gini_threshold", 0.5]
          - values: ["--method", "gic"]
          # - values: ["--method", "ica"]
      - id: fcps
        name: "fcps"
        software_environment: "fcps"
        repository:
          url: https://github.com/imallona/clustbench_fcps
          commit: 272fa5f
        parameters:
          # - values: ["--method", "FCPS_AdaptiveDensityPeak"] # not in conda
          - values: ["--method", "FCPS_Minimax"]
          - values: ["--method", "FCPS_MinEnergy"]
          # - values: ["--method", "FCPS_HDBSCAN_2"]
          # - values: ["--method", "FCPS_HDBSCAN_4"]
          # - values: ["--method", "FCPS_HDBSCAN_8"]
          # - values: ["--method", "FCPS_Diana"]
          # - values: ["--method", "FCPS_Fanny"]
          # - values: ["--method", "FCPS_Hardcl"]
          # - values: ["--method", "FCPS_Softcl"]
          # - values: ["--method", "FCPS_Clara"]
          # - values: ["--method", "FCPS_PAM"]
    inputs:
      - entries:
          - data.matrix
          - data.true_labels
    outputs:
      - id: clustering.predicted_ks_range
        path: "{input}/{stage}/{module}/{params}/{dataset}_ks_range.labels.gz"

  - id: metrics
    modules:
      - id: partition_metrics
        name: "clustbench partition metrics"
        software_environment: "clustbench"
        repository:
          url: https://github.com/imallona/clustbench_metrics
          commit: 9132d45
        parameters:
          - values: ["--metric", "normalized_clustering_accuracy"]
          - values: ["--metric", "adjusted_fm_score"]
          # - values: ["--metric", "adjusted_mi_score"]
          # - values: ["--metric", "adjusted_rand_score"]
          # - values: ["--metric", "fm_score"]
          # - values: ["--metric", "mi_score"]
          # - values: ["--metric", "normalized_clustering_accuracy"]
          # - values: ["--metric", "normalized_mi_score"]
          # - values: ["--metric", "normalized_pivoted_accuracy"]
          # - values: ["--metric", "pair_sets_index"]
          # - values: ["--metric", "rand_score"]
    inputs:
      - entries:
          - clustering.predicted_ks_range
          - data.true_labels
    outputs:
      - id: metrics.scores
        path: "{input}/{stage}/{module}/{params}/{dataset}.scores.gz"