Detail — Mlp

Metrics

• Accuracy: 95.80%
• Macro F1: 95.69%
• Weighted F1: 95.78%
• Best Val Acc: 97.38%

Config (JSON)

{
  "tag": "wide_fast_b512",
  "layers": [
    768,
    384
  ],
  "dropouts": [
    0.12,
    0.22
  ],
  "input_dropout": 0.05,
  "activation": "relu",
  "optimizer": "adam",
  "lr": 0.0014,
  "epochs": 80,
  "batch": 512,
  "use_smote": true,
  "test_size": 0.15,
  "val_size": 0.15,
  "early_stop_patience": 10,
  "reduce_lr": true,
  "rlr_factor": 0.5,
  "rlr_patience": 4
}

Source MD

md/mlp_wide_fast_b512.md