Detail — Mlp

Metrics

• Accuracy: 94.56%
• Macro F1: 94.29%
• Weighted F1: 94.51%
• Best Val Acc: 94.67%

Config (JSON)

{
  "tag": "sgd_conservative",
  "layers": [
    640,
    320,
    160
  ],
  "dropouts": [
    0.12,
    0.2,
    0.25
  ],
  "input_dropout": 0.05,
  "activation": "relu",
  "optimizer": "sgd",
  "lr": 0.008,
  "epochs": 180,
  "batch": 256,
  "use_smote": false,
  "test_size": 0.15,
  "val_size": 0.15,
  "early_stop_patience": 20,
  "reduce_lr": true,
  "rlr_factor": 0.5,
  "rlr_patience": 8,
  "rlr_min_lr": 5e-05
}

Source MD

md/mlp_sgd_conservative.md