mismm.md
In mildsvm: Multiple-Instance Learning with Support Vector Machines

mismm() works for data-frame-like inputs

Code
  bag_preds
Output
  # A tibble: 8 x 4
  # Groups:   bag_name [8]
    bag_label bag_name .pred_class  .pred
        <dbl> <chr>    <fct>        <dbl>
  1         0 bag1     1            13.6 
  2         0 bag2     1           164.  
  3         0 bag3     1            23.7 
  4         1 bag4     1            47.1 
  5         0 bag5     0           -33.2 
  6         0 bag6     0           -31.4 
  7         0 bag7     1             8.95
  8         1 bag8     1           643.

Code
  bag_preds
Output
  # A tibble: 8 x 4
  # Groups:   bag_name [8]
    bag_label bag_name .pred_class .pred
        <dbl> <chr>    <fct>       <dbl>
  1         0 bag1     1           0.826
  2         0 bag2     1           0.780
  3         0 bag3     1           0.705
  4         1 bag4     1           1.00 
  5         0 bag5     1           0.885
  6         0 bag6     1           0.763
  7         0 bag7     1           0.706
  8         1 bag8     1           0.809

Code
  bag_preds
Output
  # A tibble: 8 x 4
  # Groups:   bag_name [8]
    bag_label bag_name .pred_class  .pred
        <dbl> <chr>    <fct>        <dbl>
  1         0 bag1     0           -0.731
  2         0 bag2     0           -0.821
  3         0 bag3     0           -0.768
  4         1 bag4     0           -0.617
  5         0 bag5     0           -0.809
  6         0 bag6     0           -0.779
  7         0 bag7     0           -0.800
  8         1 bag8     0           -0.638

`mismm()` value returns make sense

Code
  models <- list(`mildata-heur` = mismm(df, method = "heuristic"), `mildata-mip` = mismm(
    df, method = "mip", control = list(nystrom_args = list(m = 10))),
  `mildata-qp` = mismm(df, method = "qp-heuristic"), xy = mismm(x = as.data.frame(
    df[, 4:6]), y = df$bag_label, bags = df$bag_name, instances = df$
  instance_name), formula = mismm(mild(bag_label, bag_name, instance_name) ~ .,
  data = df), `no-scale-heur` = mismm(df, method = "heuristic", control = list(
    scale = FALSE)), `no-scale-mip` = mismm(df, method = "mip", control = list(
    scale = FALSE, nystrom_args = list(m = 10))), `no-scale-qp` = mismm(df,
    method = "qp-heuristic", control = list(scale = FALSE)), `no-weights` = mismm(
    df, method = "heuristic", weights = FALSE)) %>% suppressWarnings() %>%
    suppressMessages()
  print(lapply(models, names))
Output
  $`mildata-heur`
   [1] "ksvm_fit"        "call_type"       "x"               "features"       
   [5] "levels"          "cost"            "sigma"           "weights"        
   [9] "kernel"          "kernel_param"    "repr_inst"       "n_step"         
  [13] "useful_inst_idx" "inst_order"      "x_scale"         "bag_name"       
  [17] "instance_name"

  $`mildata-mip`
   [1] "gurobi_fit"    "kfm_fit"       "call_type"     "features"     
   [5] "levels"        "cost"          "sigma"         "weights"      
   [9] "kernel"        "kernel_param"  "x_scale"       "bag_name"     
  [13] "instance_name"

  $`mildata-qp`
   [1] "gurobi_fit"    "call_type"     "x"             "features"     
   [5] "levels"        "cost"          "sigma"         "weights"      
   [9] "kernel"        "kernel_param"  "repr_inst"     "n_step"       
  [13] "x_scale"       "bag_name"      "instance_name"

  $xy
   [1] "ksvm_fit"        "call_type"       "x"               "features"       
   [5] "levels"          "cost"            "sigma"           "weights"        
   [9] "kernel"          "kernel_param"    "repr_inst"       "n_step"         
  [13] "useful_inst_idx" "inst_order"      "x_scale"

  $formula
   [1] "ksvm_fit"        "call_type"       "x"               "features"       
   [5] "levels"          "cost"            "sigma"           "weights"        
   [9] "kernel"          "kernel_param"    "repr_inst"       "n_step"         
  [13] "useful_inst_idx" "inst_order"      "x_scale"         "formula"        
  [17] "bag_name"        "instance_name"

  $`no-scale-heur`
   [1] "ksvm_fit"        "call_type"       "x"               "features"       
   [5] "levels"          "cost"            "sigma"           "weights"        
   [9] "kernel"          "kernel_param"    "repr_inst"       "n_step"         
  [13] "useful_inst_idx" "inst_order"      "bag_name"        "instance_name"

  $`no-scale-mip`
   [1] "gurobi_fit"    "kfm_fit"       "call_type"     "features"     
   [5] "levels"        "cost"          "sigma"         "weights"      
   [9] "kernel"        "kernel_param"  "bag_name"      "instance_name"

  $`no-scale-qp`
   [1] "gurobi_fit"    "call_type"     "x"             "features"     
   [5] "levels"        "cost"          "sigma"         "weights"      
   [9] "kernel"        "kernel_param"  "repr_inst"     "n_step"       
  [13] "bag_name"      "instance_name"

  $`no-weights`
   [1] "ksvm_fit"        "call_type"       "x"               "features"       
   [5] "levels"          "cost"            "sigma"           "kernel"         
   [9] "kernel_param"    "repr_inst"       "n_step"          "useful_inst_idx"
  [13] "inst_order"      "x_scale"         "bag_name"        "instance_name"

Code
  print(models)
Output
  $`mildata-heur`
  An mismm object called with mismm.mild_df

  Parameters: 
    method: heuristic 
    kernel: kme w/ radial  (sigma = 0.3333333) 
    cost: 1 
    scale: TRUE 
    weights: ('0' = 0.166666666666667, '1' = 1)

  Model info: 
    Features: chr [1:3] "X1" "X2" "X3"
    Number of iterations: 2


  $`mildata-mip`
  An mismm object called with mismm.mild_df

  Parameters: 
    method: mip 
    kernel: kme w/ radial  (sigma = 0.3333333) 
    cost: 1 
    scale: TRUE 
    weights: ('0' = 0.166666666666667, '1' = 1)

  Model info: 
    Features: chr [1:3] "X1" "X2" "X3"
    Gap to optimality: 0


  $`mildata-qp`
  An mismm object called with mismm.mild_df

  Parameters: 
    method: qp-heuristic 
    kernel: kme w/ radial  (sigma = 0.3333333) 
    cost: 1 
    scale: TRUE 
    weights: ('0' = 0.166666666666667, '1' = 1)

  Model info: 
    Features: chr [1:3] "X1" "X2" "X3"
    Number of iterations: 1


  $xy
  An mismm object called with mismm.default

  Parameters: 
    method: heuristic 
    kernel: kme w/ radial  (sigma = 0.3333333) 
    cost: 1 
    scale: TRUE 
    weights: ('0' = 0.166666666666667, '1' = 1)

  Model info: 
    Features: chr [1:3] "X1" "X2" "X3"
    Number of iterations: 2


  $formula
  An mismm object called with mismm.formula

  Parameters: 
    method: heuristic 
    kernel: kme w/ radial  (sigma = 0.3333333) 
    cost: 1 
    scale: TRUE 
    weights: ('0' = 0.166666666666667, '1' = 1)

  Model info: 
    Features: chr [1:3] "X1" "X2" "X3"
    Number of iterations: 2


  $`no-scale-heur`
  An mismm object called with mismm.mild_df

  Parameters: 
    method: heuristic 
    kernel: kme w/ radial  (sigma = 0.3333333) 
    cost: 1 
    scale: FALSE 
    weights: ('0' = 0.166666666666667, '1' = 1)

  Model info: 
    Features: chr [1:3] "X1" "X2" "X3"
    Number of iterations: 2


  $`no-scale-mip`
  An mismm object called with mismm.mild_df

  Parameters: 
    method: mip 
    kernel: kme w/ radial  (sigma = 0.3333333) 
    cost: 1 
    scale: FALSE 
    weights: ('0' = 0.166666666666667, '1' = 1)

  Model info: 
    Features: chr [1:3] "X1" "X2" "X3"
    Gap to optimality: 0


  $`no-scale-qp`
  An mismm object called with mismm.mild_df

  Parameters: 
    method: qp-heuristic 
    kernel: kme w/ radial  (sigma = 0.3333333) 
    cost: 1 
    scale: FALSE 
    weights: ('0' = 0.166666666666667, '1' = 1)

  Model info: 
    Features: chr [1:3] "X1" "X2" "X3"
    Number of iterations: 1


  $`no-weights`
  An mismm object called with mismm.mild_df

  Parameters: 
    method: heuristic 
    kernel: kme w/ radial  (sigma = 0.3333333) 
    cost: 1 
    scale: TRUE 
    weights: FALSE

  Model info: 
    Features: chr [1:3] "X1" "X2" "X3"
    Number of iterations: 2