inst/experiments/exp01_seqLen10_nSeq300_amylogram_encoding_infogain.R

In jakubkala/QuiPTsim:

library(QuiPTsim)
library(drake)

plan <- drake_plan(

  ###### data frame containing datasets' details and their paths to RDS files
  df = read.csv("~/projects/QuiPTsim-data/reduced_alph_enc_amylogram_encoding_unigram/amylogram_encoding.csv"),

  numSeq = 300,
  fraction = 0.5,
  size = 3,
  ###### selected paths
  paths_motifs1 = sample(paste0("~/projects/QuiPTsim-data/reduced_alph_enc_amylogram_encoding_unigram/",
                sapply(strsplit(x = df[df$l_seq==10 & df$n_motifs==1,"path"], split = "/"), function(x) x[[3]])),
                size = size),

  paths_motifs2 = sample(paste0("~/projects/QuiPTsim-data/reduced_alph_enc_amylogram_encoding_unigram/",
                                sapply(strsplit(x = df[df$l_seq==10 & df$n_motifs==2,"path"], split = "/"), function(x) x[[3]])),
                         size = size),

  paths_motifs3 = sample(paste0("~/projects/QuiPTsim-data/reduced_alph_enc_amylogram_encoding_unigram/",
                                sapply(strsplit(x = df[df$l_seq==10 & df$n_motifs==3,"path"], split = "/"), function(x) x[[3]])),
                         size = size),
  paths = c(paths_motifs1, paths_motifs2, paths_motifs3),

  output_prefix = "~/experiment-results/exp01-seqLen10-nSeq300-amylogram-encoding-infogain/result_",

  ###### details of models used in ranking comparison
  models_details = list(
    list(model = "lm",
         param_name = "lambda",
         param_value = NULL),
    list(model = "knn",
         param_name = "neighbors",
         param_value = 2^(0:4)),
    list(model = "rf",
         param_name = "num.trees",
         param_value = c(500, 1000)),
    list(model = "naive bayes",
         param_name = "laplace",
         param_value = 0)
  ),

  ###### Validation scheme
  validation_scheme = list(type = "cv",
                           folds = 5,
                           n_kmers = c(2:128, 2^(8:12)),
                           cv_reps = 1,
                           models_details = models_details),

  validation_scheme_nonranking = list(type = "cv",
                                      folds = 5,
                                      cv_reps = 1,
                                      models_details = models_details),


  ###### Validation scheme
  filter_names = c("QuiPT",
                   "Chi-squared",
                   "FCBF",
                   "infogain", "gainratio", "symuncert",
                   "MRMR", "JMI", "JMIM", "DISR", "CMIM", "NJMIM"),

  # ranking_QuiPT = filter_rankings(paths, output_prefix, "QuiPT", numSeq, fraction, validation_scheme),
  # ranking_Chi = filter_rankings(paths, output_prefix, "Chi-squared", numSeq, fraction, validation_scheme),
  ranking_infogain = filter_rankings(paths, output_prefix, "infogain", numSeq, fraction, validation_scheme),
  # ranking_gainratio = filter_rankings(paths, output_prefix, "gainratio", numSeq, fraction, validation_scheme),
  # ranking_symuncert = filter_rankings(paths, output_prefix, "symuncert", numSeq, fraction, validation_scheme),
  # ranking_MRMR = filter_rankings(paths, output_prefix, "MRMR", numSeq, fraction, validation_scheme),
  # ranking_JMI = filter_rankings(paths, output_prefix, "JMI", numSeq, fraction, validation_scheme),
  # ranking_JMIM = filter_rankings(paths, output_prefix, "JMIM", numSeq, fraction, validation_scheme),
  # ranking_DISR = filter_rankings(paths, output_prefix, "DISR", numSeq, fraction, validation_scheme),
  # ranking_NJMIM = filter_rankings(paths, output_prefix, "NJMIM", numSeq, fraction, validation_scheme),


  # thresholds = c(0.01, 0.05),

  # nonranking_QuiPT = filter_nonrankings(paths, output_prefix, "QuiPT", numSeq, fraction, validation_scheme_nonranking,
  #                                        thresholds),
  # nonranking_Chi = filter_nonrankings(paths, output_prefix, "Chi-squared", numSeq, fraction, validation_scheme_nonranking,
  #                                      thresholds),

  # nonranking_gainratio = filter_nonrankings(paths, output_prefix, "gainratio", numSeq, fraction, validation_scheme_nonranking),
  # nonranking_infogain = filter_nonrankings(paths, output_prefix, "infogain", numSeq, fraction, validation_scheme_nonranking),
  # nonranking_symuncert = filter_nonrankings(paths, output_prefix, "symuncert", numSeq, fraction, validation_scheme_nonranking),

  # nonranking_FCBF = filter_nonrankings(paths, output_prefix, "FCBF", numSeq, fraction, validation_scheme_nonranking)
)


cache <- new_cache("exp01_seqLen10_nSeq300_amylogram_encoding_infogain")
make(
  plan,
  parallelism = "future",
  jobs = 8,
  log_make = "exp01_seqLen10_nSeq300_amylogram_encoding_infogain.log",
  cache = cache,
  seed = 42
)