Models & scores

[veniversum]

• MLP

  • Dense(128, ReLU), Dense(2, softmax), categorized outputs; batch size 128

  • 0.84665 score 5 fold cv

  • Dense(256, ReLU), Dense(2, softmax), categorized outputs; batch size 128

  • 0.84615 score 5 fold cv

  • Dense(512, ReLU, 0.5 dropout, 0.01 L2 reg), Dense(2, softmax), categorized outputs;

    • batch size 256
    • 0.03 L2 reg

  • batch size 128

  • batch size 64

  • batch size 32

• RandomForest (sklearn.ensemble.RandomForestClassifier)

  • 10 trees, default params
    • 0.77601 score 5 fold cv
  • 50 trees, default params
  • 0.8193505 score 5 fold cv
  • 100 trees, default params
  • 0.825700 score 5 fold cv
  • 200 trees, RandomForestClassifier(n_estimators=200, max_features='sqrt', criterion='gini', min_samples_leaf=2, min_samples_split=178)
  • 0.823749 score 5 fold cv
  • 200 trees, RandomForestClassifier(n_estimators=200, max_features='sqrt', criterion='gini', min_samples_leaf=2, min_samples_split=135)
  • 0.826700 score 5 fold cv

[veniversum]

[veniversum]

word magic 1 word magic 2 128 0.9 flatten 128 0.9 1

128 0.01 0.5 flatten 128 0.01 0.5 1

128 0.001 0.7 flatten 128 0.001 0.7 1