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

`mior()` has reasonable performance

Code
  print(roc$auc)
Output
  Multi-class area under the curve: 0.6775
Code
  print(mzoe)
Output
  [1] 0.8466667
Code
  print(mae)
Output
  [1] 1.373333
Code
  print(table(true, pred))
Output
      pred
  true  1  3
     1 18 30
     2 24 24
     3 49  5

Code
  print(roc$auc)
Output
  Multi-class area under the curve: 0.5973
Code
  print(mzoe)
Output
  [1] 0.76
Code
  print(mae)
Output
  [1] 1.173333
Code
  print(table(true, pred))
Output
      pred
  true  1  3
     1 22 30
     2 21 31
     3 32 14

Code
  print(roc$auc)
Output
  Multi-class area under the curve: 0.6057
Code
  print(mzoe)
Output
  [1] 0.7
Code
  print(mae)
Output
  [1] 0.7733333
Code
  print(table(true, pred))
Output
      pred
  true  1  2
     1 22 26
     2 25 23
     3 11 43

Code
  print(roc$auc)
Output
  Multi-class area under the curve: 0.5268
Code
  print(mzoe)
Output
  [1] 0.68
Code
  print(mae)
Output
  [1] 0.7733333
Code
  print(table(true, pred))
Output
      pred
  true  1  2
     1 20 32
     2 24 28
     3 14 32

`mior()` works for data-frame-like inputs

Code
  predict(mdl2, new_data = df1, type = "class", layer = "bag")
Output
  # A tibble: 450 x 1
     .pred_class
     <fct>      
   1 1          
   2 1          
   3 2          
   4 1          
   5 2          
   6 1          
   7 2          
   8 1          
   9 1          
  10 1          
  # ... with 440 more rows
Code
  predict(mdl2, new_data = df1, type = "class", layer = "instance")
Output
  # A tibble: 450 x 1
     .pred_class
     <fct>      
   1 1          
   2 1          
   3 1          
   4 1          
   5 1          
   6 1          
   7 1          
   8 1          
   9 1          
  10 1          
  # ... with 440 more rows
Code
  predict(mdl2, new_data = df1, type = "raw", layer = "bag")
Output
  # A tibble: 450 x 1
      .pred
      <dbl>
   1  1.84 
   2  0.455
   3 -1.56 
   4  0.566
   5 -2.61 
   6  0.622
   7 -1.13 
   8  2.78 
   9  0.403
  10  2.48 
  # ... with 440 more rows
Code
  predict(mdl2, new_data = df1, type = "raw", layer = "instance")
Output
  # A tibble: 450 x 1
      .pred
      <dbl>
   1  1.84 
   2 10.9  
   3  3.44 
   4  0.566
   5  2.38 
   6  5.63 
   7  2.85 
   8  2.78 
   9  0.681
  10  2.48 
  # ... with 440 more rows

`mior()` value returns make sense

Code
  models <- list(xy = mior(x = df1[, 4:6], y = df1$bag_label, bags = df1$bag_name,
  method = "qp-heuristic", weights = NULL), formula = mior(mi(bag_label, bag_name) ~
    V1 + V2, method = "qp-heuristic", data = df1, weights = NULL), mi_df = mior(
    as_mi_df(df1, instance_label = NULL)), `no-scale` = mior(x = df1[, 4:6], y = df1$
    bag_label, bags = df1$bag_name, method = "qp-heuristic", weights = NULL,
  control = list(scale = FALSE))) %>% suppressWarnings() %>% suppressMessages()
  print(lapply(models, names))
Output
  $xy
   [1] "gurobi_fit" "call_type"  "features"   "levels"     "cost"      
   [6] "cost_eta"   "kernel"     "repr_inst"  "n_step"     "x_scale"

  $formula
   [1] "gurobi_fit" "call_type"  "features"   "levels"     "cost"      
   [6] "cost_eta"   "kernel"     "repr_inst"  "n_step"     "x_scale"   
  [11] "formula"    "bag_name"

  $mi_df
   [1] "gurobi_fit" "call_type"  "features"   "levels"     "cost"      
   [6] "cost_eta"   "kernel"     "repr_inst"  "n_step"     "x_scale"   
  [11] "bag_name"

  $`no-scale`
  [1] "gurobi_fit" "call_type"  "features"   "levels"     "cost"      
  [6] "cost_eta"   "kernel"     "repr_inst"  "n_step"

Code
  print(models)
Output
  $xy
  An mior object called with mior.default

  Parameters: 
    method: qp-heuristic 
    kernel: linear  
    cost: 1 
    cost_eta: 1 
    scale: TRUE 
    weights: FALSE

  Model info: 
    Levels of `y`: chr [1:3] "1" "2" "3"
    Features: chr [1:3] "V1" "V2" "V3"
    Number of iterations: 4 
    Gap to optimality:


  $formula
  An mior object called with mior.formula

  Parameters: 
    method: qp-heuristic 
    kernel: linear  
    cost: 1 
    cost_eta: 1 
    scale: TRUE 
    weights: FALSE

  Model info: 
    Levels of `y`: chr [1:3] "1" "2" "3"
    Features: chr [1:2] "V1" "V2"
    Number of iterations: 5 
    Gap to optimality:


  $mi_df
  An mior object called with mior.mi_df

  Parameters: 
    method: qp-heuristic 
    kernel: linear  
    cost: 1 
    cost_eta: 1 
    scale: TRUE 
    weights: FALSE

  Model info: 
    Levels of `y`: chr [1:3] "1" "2" "3"
    Features: chr [1:4] "repr" "V1" "V2" "V3"
    Number of iterations: 4 
    Gap to optimality:


  $`no-scale`
  An mior object called with mior.default

  Parameters: 
    method: qp-heuristic 
    kernel: linear  
    cost: 1 
    cost_eta: 1 
    scale: FALSE 
    weights: FALSE

  Model info: 
    Levels of `y`: chr [1:3] "1" "2" "3"
    Features: chr [1:3] "V1" "V2" "V3"
    Number of iterations: 4 
    Gap to optimality: