Skip to content

Latest commit

 

History

History
37 lines (33 loc) · 2.01 KB

File metadata and controls

37 lines (33 loc) · 2.01 KB

Logs

Trials for data models are recorded here.

Data

718887 battles; Absolute trophy diff < 1000; Trophy data scaled;

ANN

  • Trial 1:
    • Hidden Layers:
      • 1: neurons=144, activation="relu", kernel_regularizer=L1(0.0001)
      • 2: neurons=144, activation="relu", kernel_regularizer=L1(0.0001)
      • 3: (Output): activation="sigmoid", kernel_regularizer=L1(0.0001)
    • Optimizer: SGD, learning_rate=0.0025, momentum=0.95, nesterov=True
    • Training: batch_size=64
    • Results: Stopped at 31/512 epochs; Validation Accuracy is 0.583
  • Trial 2:
    • Hidden Layers:
      • 1: neurons=144, activation="relu", kernel_regularizer=L1(0.0001), kernel_initializer="he_uniform"
      • 2: neurons=144, activation="relu", kernel_regularizer=L1(0.0001), kernel_initializer="he_uniform"
      • 3: (Output): activation="sigmoid", kernel_regularizer=L1(0.0001)
    • Optimizer: Adam, learning_rate=0.00025
    • Training: batch_size=128
    • Results: Stopped at 85/512 epochs' Validation Accuracy is 0.592
  • Trial 3:
    • Hidden Layers:
      • 1: neurons=144, activation="relu", kernel_regularizer=L1(0.0001), kernel_initializer="he_uniform"
      • 2: neurons=72, activation="relu", kernel_regularizer=L1(0.0001), kernel_initializer="he_uniform"
      • 3: neurons=36, activation="relu", kernel_regularizer=L1(0.0001), kernel_initializer="he_uniform"
      • 4: neurons=36, activation="relu", kernel_regularizer=L1(0.0001), kernel_initializer="he_uniform"
      • 5: neurons=36, activation="relu", kernel_regularizer=L1(0.0001), kernel_initializer="he_uniform"
      • 6: neurons=36, activation="relu", kernel_regularizer=L1(0.0001), kernel_initializer="he_uniform"
      • 7: (Output) activation="sigmoid", kernel_regularizer=L1(0.0001)
    • Optimizer: Adam, learning_rate=0.00025
    • Training: batch_size=196
    • Results: Stopped at 38/512 epochs' Validation Accuracy is 0.597